notebooks/01_project_refresh.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "db8d4f31-c5a7-4e8f-bd61-f39fd0fef7a5",
   "metadata": {},
   "source": [
    "# Japanese Character Recognition"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "46e44108-9728-404d-a911-13b54997b5c3",
   "metadata": {},
   "source": [
    "This notebook is a revisit of an earlier project that leveraged neural networks for Japanese character recognition that implements better techniques. The previous iteration was rushed to check all the boxes for an MVP (while having the added burdon of building an app and a self generated dataset) and this document compares various improvements to the earier approaches with a more targetted and suitable approach to the task.\n",
    "\n",
    "Unlike most other character recognition projects, this one uses character data that is entirely digitial by recording the strokes and component points logged while interacting with the canvas of an Android App. This provides several advantages that more traditional datasets do not offer:\n",
    "* More compact storage of the data when stored as vectors of x and y coordinates\n",
    "* Additional stroke order information is gained which can either be used directly in RNNs or represented by color shifts augmenting the inputs of a CNN.\n",
    "* Contains additional stroke direction information that can derive additional features within an RNN.\n",
    "\n",
    "This notebook will focus on correctly identifying characters from the standard hiragana monographs and digraphs with and without diacritic with 108 possible classes in total. In the future, there may be subsequent notebooks that expand to the Katakana and Kanji character sets, but those will less exploratory in nature and take on a more narrow approach with a more defined strategy."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d64454c2-1157-449e-9085-e597849094fb",
   "metadata": {},
   "source": [
    "## Setup"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2934036c-20db-4023-b873-c517273f5908",
   "metadata": {},
   "source": [
    "The majority of the code is contained within the src directory with the exception of explicit PyTorch model architecture definitions and any methods or classes that are unlikely to be firmly established for reuse in any subsequent notebooks."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "d943083a-b2da-46ae-a495-9cb6b8a8608f",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import random\n",
    "import sys\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "from sklearn.dummy import DummyClassifier\n",
    "from sklearn.metrics import accuracy_score, balanced_accuracy_score\n",
    "import pandas as pd\n",
    "\n",
    "PWD = os.getcwd()\n",
    "sys.path.append(os.path.dirname(PWD))\n",
    "from src import StrokeDataPaths, StrokeDataset, ToBSplines, ArrayToTensor, ExtractAngles, EmptyStrokePadder, CharacterTrainer, StrokesToPil, IncorrectCharacters, InputNormalizer, InputMinMaxTransformer, InputRecenter, StrokeDatasetAggregator, StrokeExtractAbsolute, InputGaussianNoise\n",
    "\n",
    "#%matplotlib widget\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "88704eec-218a-453f-b106-ebf23180943c",
   "metadata": {},
   "outputs": [],
   "source": [
    "from torchvision.transforms import v2 as transforms\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as f"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a012546b-7007-4ea5-9f1b-611bf764a9d7",
   "metadata": {},
   "source": [
    "Present in the following cell are some methods to preserve reproduceability, but it is rendered mostly inneffective because a complete approach would break compatibility with pytorch's VGG16 implementation used in the transfer learning CNN. PyTorch's AdaptiveAvgPool2d layer currently does not support deterministic operations, so the code to force it on the GPU is disabled in order to optimize GPU performance instead of trying to meet a constraint that cannot be effectively met for all the presented architectures."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "8f3e2b23-1a54-4141-a2e0-7b71ef4cb627",
   "metadata": {},
   "outputs": [],
   "source": [
    "SEED = 1111\n",
    "torch.manual_seed(SEED)\n",
    "random.seed(SEED)\n",
    "np.random.seed(SEED)\n",
    "# Determinism is not supported by one of the layers in the VGG16 model\n",
    "#torch.use_deterministic_algorithms(True, warn_only=True)\n",
    "#os.environ[\"CUBLAS_WORKSPACE_CONFIG\"] = \":4096:8\""
   ]
  },
  {
   "cell_type": "markdown",
   "id": "90617262-8cdc-428e-824d-deb8d08de572",
   "metadata": {},
   "source": [
    "The path referring to the raw data currently points to a json file based dataset that is not stored on the repo, but it will be made available on HuggingFace at some time in the future."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "156fddba-7bc9-413f-9511-debf9a405903",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Stroring the common paths in variables\n",
    "RAW_PATH = \"../data/choubenkyo_data\"\n",
    "CLASS_PATH = \"../choubenkyo_kivy_app/app/draw_screen/characters\"\n",
    "# While different characters contain different stroke counts, in order for them to easily work\n",
    "# as tensors with fully random shuffling, they will need to be padded to a set maximum.\n",
    "stroke_count = 10\n",
    "batch_size = 20"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "72aa730c-ea95-4cf6-845e-21c7ad92615b",
   "metadata": {},
   "source": [
    "To perform some basic EDA on this dataset, this cell instantiates two custom objects that simplify the loading of the raw data, labels, and additional metadata within the dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "ec0234fb-7d77-44b5-86d5-77a170a203df",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Custom class that serves as a bridge between the Dataset object and the file system. It manages paths and extracts information contained within structured filenames.\n",
    "data_paths = StrokeDataPaths(RAW_PATH, CLASS_PATH)\n",
    "# A custom dataset object that also allows the use of constraints based on a character's stroke count.\n",
    "full_dataset = StrokeDataset(data_paths.get_data())"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "06614431-edce-4483-aef0-edd546f41a3b",
   "metadata": {},
   "source": [
    "## Exploration"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9910c208-3b27-49ca-a8bd-a247d35d7df9",
   "metadata": {},
   "source": [
    "A few notebook-specific functions to reduce boilerplate code in plotting."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "8fde9dd9-417c-4f86-8336-4103bf31cec3",
   "metadata": {},
   "outputs": [],
   "source": [
    "def common_plot_setup(figure=None, axes=None, palette=None, font=None, figsize=(15, 10), style=None, rows=1, cols=1):\n",
    "    if style:\n",
    "        plt.style.use(style)\n",
    "    if font:\n",
    "        plt.rcParams['font.family'] = font\n",
    "    if palette:\n",
    "        sns.set_palette(palette)\n",
    "    if figure and axes:\n",
    "        fig, ax = figure, axes\n",
    "    else:\n",
    "        fig, ax = plt.subplots(nrows=rows, ncols=cols, figsize=figsize)\n",
    "    return fig, ax\n",
    "\n",
    "def colors_from_values(values, palette_name):\n",
    "    \"\"\"Creates a palette that changes coloring of bar plots to vary by their y value\"\"\"\n",
    "\n",
    "    # normalize the values to range [0, 1]\n",
    "    normalized = (values - min(values)) / (max(values) - min(values))\n",
    "    # convert to indices\n",
    "    indices = np.round(normalized * (len(values) - 1)).astype(np.int32)\n",
    "    # use the indices to get the colors\n",
    "    palette = sns.color_palette(palette_name, len(values))\n",
    "    return np.array(palette).take(indices, axis=0).tolist()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "244348cb-b1d5-405f-b0c3-09880fa8efba",
   "metadata": {},
   "source": [
    "### Dataset Overview"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6c6b570d-d788-443d-80f4-1d84af9f3965",
   "metadata": {},
   "source": [
    "The custom dataset used in this notebook will be significantly improved from its first iteration. When the MVP was built, there wasn't any time available to spend to detect incorrect inputs that only serve to hinder the accuracy on the model.\n",
    "\n",
    "Because the Japanese writing system has a defined structure regarding stroke order, the optimal application would be best served adhering to that structure especially when considering the idea of real time stroke by stroke character prediction that could have to deal with thousands of potential classes on a partially finished character. Generalizing to many different handwriting styles would still be advantageous, but the overall steps in drawing should be consistent.\n",
    "\n",
    "The following code shows the overall size of the dataset and a table containing the amount of available obeservations for each character."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "b40899c1-53cb-43d7-a563-173bae2ba30d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total dataset size: 6716 drawn hiragana characters\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>count</th>\n",
       "      <th>label</th>\n",
       "      <th>char_id</th>\n",
       "      <th>character</th>\n",
       "      <th>romanization</th>\n",
       "      <th>stroke_count</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>label</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>60</td>\n",
       "      <td>hir_a</td>\n",
       "      <td>a</td>\n",
       "      <td>あ</td>\n",
       "      <td>a</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>60</td>\n",
       "      <td>hir_i</td>\n",
       "      <td>i</td>\n",
       "      <td>い</td>\n",
       "      <td>i</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>58</td>\n",
       "      <td>hir_u</td>\n",
       "      <td>u</td>\n",
       "      <td>う</td>\n",
       "      <td>u</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>59</td>\n",
       "      <td>hir_e</td>\n",
       "      <td>e</td>\n",
       "      <td>え</td>\n",
       "      <td>e</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>42</td>\n",
       "      <td>hir_o</td>\n",
       "      <td>o</td>\n",
       "      <td>お</td>\n",
       "      <td>o</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>103</th>\n",
       "      <td>62</td>\n",
       "      <td>hir_byu</td>\n",
       "      <td>byu</td>\n",
       "      <td>びゅ</td>\n",
       "      <td>byu</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>104</th>\n",
       "      <td>46</td>\n",
       "      <td>hir_byo</td>\n",
       "      <td>byo</td>\n",
       "      <td>びょ</td>\n",
       "      <td>byo</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>105</th>\n",
       "      <td>51</td>\n",
       "      <td>hir_pya</td>\n",
       "      <td>pya</td>\n",
       "      <td>ぴゃ</td>\n",
       "      <td>pya</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>106</th>\n",
       "      <td>58</td>\n",
       "      <td>hir_pyu</td>\n",
       "      <td>pyu</td>\n",
       "      <td>ぴゅ</td>\n",
       "      <td>pyu</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>107</th>\n",
       "      <td>56</td>\n",
       "      <td>hir_pyo</td>\n",
       "      <td>pyo</td>\n",
       "      <td>ぴょ</td>\n",
       "      <td>pyo</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>108 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       count    label char_id character romanization stroke_count\n",
       "label                                                            \n",
       "0         60    hir_a       a         あ            a            3\n",
       "1         60    hir_i       i         い            i            2\n",
       "2         58    hir_u       u         う            u            2\n",
       "3         59    hir_e       e         え            e            2\n",
       "4         42    hir_o       o         お            o            3\n",
       "...      ...      ...     ...       ...          ...          ...\n",
       "103       62  hir_byu     byu        びゅ          byu            5\n",
       "104       46  hir_byo     byo        びょ          byo            5\n",
       "105       51  hir_pya     pya        ぴゃ          pya            5\n",
       "106       58  hir_pyu     pyu        ぴゅ          pyu            4\n",
       "107       56  hir_pyo     pyo        ぴょ          pyo            4\n",
       "\n",
       "[108 rows x 6 columns]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "counts = pd.DataFrame(np.unique(full_dataset.labels, return_counts=True), index=[\"label\", \"count\"]).T.set_index(\"label\")\n",
    "counts = counts.join(data_paths.get_classes(), how=\"left\")\n",
    "print(\"Total dataset size:\", counts[\"count\"].sum(), \"drawn hiragana characters\")\n",
    "counts"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b440540c-b0a0-45e8-a40e-d01d27da00f7",
   "metadata": {},
   "source": [
    "Since the app was designed to randomly present characters for the capture of user input, there is moderate coverage of all classes. As expected, differences in counts follow a roughly normal distribution. Training would likely benefit slightly from some class balancing, but it would be more efficient to focus on integrating that more thoughtfully in a future notebook since the options presented here serve only as a comparative starting point. The evaluation metric will be the model's balanced accuracy score to better offset any advantages if the model were to favor the dominant classes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "71d13f8b-ea64-4079-9d93-4ff35c3a022a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1500x1000 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = common_plot_setup(style=\"fivethirtyeight\", font=\"Noto Sans\")\n",
    "sns.barplot(x = \"character\", y=\"count\", data=counts, ax=ax, hue=\"character\", palette=colors_from_values(counts[\"count\"], \"Spectral\"), legend=False)\n",
    "plt.xticks(fontname = \"Noto Sans JP\")\n",
    "[item.set_fontsize(5) for item in ax.get_xticklabels()]\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "c28d7853-c081-498b-9348-2767dc026c98",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.histplot(counts[\"count\"], bins=12)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "daccb9d0-0f83-46ea-af6a-cbcca0ea9b16",
   "metadata": {},
   "source": [
    "In addition, the stroke count is very useful to know in order to best pad all tensors for use in batches. Here we will examine a rudimentary approach that just pads them with empty data to a maximum count of 10 strokes. There are likely significant improvements to be found in the real time detection use case if instead a form of variable truncating and padding were to be used instead. The implementation warrants further thought in how exactly it could be sampled in a way that favors completed characters while maintaining a good balance between classes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "37124e69-2652-4e39-af15-b388e929204e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>class_counts</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>stroke_count</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>21</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>19</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              class_counts\n",
       "stroke_count              \n",
       "1                       11\n",
       "2                       20\n",
       "3                       24\n",
       "4                       21\n",
       "5                       19\n",
       "6                        9\n",
       "7                        3\n",
       "8                        1"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "stroke_counts = data_paths.get_data().groupby([\"label\", \"stroke_count\"]).count().groupby(\"stroke_count\").count()\n",
    "stroke_counts.columns = [\"class_counts\"]\n",
    "stroke_counts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "e473df9a-c0b7-4d26-84fe-06635ec71573",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 900x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = common_plot_setup(style=\"fivethirtyeight\", font=\"Noto Sans\", figsize=(9, 6))\n",
    "sns.barplot(x = stroke_counts.index, y=\"class_counts\", data=stroke_counts, ax=ax, hue=stroke_counts.index, palette=colors_from_values(stroke_counts[\"class_counts\"], \"Spectral\"), legend=False)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c562fa63-44ab-42d2-bb4e-3458938df928",
   "metadata": {},
   "source": [
    "### The Structure of a Single Datapoint"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6eaafb82-8f06-4f5d-9f22-4db8b2de7a52",
   "metadata": {},
   "source": [
    "The canvas capturing app encodes the data in the form of a json containing a list of sequential x and y values for every stroke. During capture, the canvas is locked to a square aspect ratio and the device display sizes are normalized so that each coordinate can only range from -1 to 1 on either axis. The custom dataset object used here simply loads the json and orders the strokes sequentially into a nested list of (strokes, datapoints, x and y values)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "10cabfeb-3c50-4984-9603-cdc5d31597cb",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "label           hir_zyu\n",
      "char_id             zyu\n",
      "character            じゅ\n",
      "romanization         ju\n",
      "stroke_count          5\n",
      "Name: 97, dtype: object\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[[[-0.6321307685640123, 0.834404706954956],\n",
       "  [-0.6408074498176575, 0.8209269046783447],\n",
       "  [-0.6662941310140822, 0.7689626216888428],\n",
       "  [-0.6943145063188341, 0.6837601661682129],\n",
       "  [-0.7131664951642355, 0.5885944366455078],\n",
       "  [-0.747886233859592, 0.3708760738372803],\n",
       "  [-0.7675962977939181, 0.22816967964172363],\n",
       "  [-0.7724891437424554, 0.07062387466430664],\n",
       "  [-0.7236234678162469, -0.19337773323059082],\n",
       "  [-0.6915084852112664, -0.2613563537597656],\n",
       "  [-0.661317937903934, -0.30898046493530273],\n",
       "  [-0.6246270404921638, -0.35156917572021484],\n",
       "  [-0.5735724833276536, -0.3908364772796631],\n",
       "  [-0.5150567822986178, -0.4151954650878906],\n",
       "  [-0.4497993654674954, -0.42128443717956543],\n",
       "  [-0.38660042815738255, -0.4132080078125],\n",
       "  [-0.34005363782246906, -0.3995366096496582],\n",
       "  [-0.30849019686381024, -0.3901848793029785],\n",
       "  [-0.2783587906095717, -0.3726654052734375],\n",
       "  [-0.26539987987942165, -0.3557295799255371],\n",
       "  [-0.17781886789533827, -0.19186925888061523]],\n",
       " [[-0.3957241111331516, 0.5361888408660889],\n",
       "  [-0.38907501432630753, 0.5295364856719971],\n",
       "  [-0.3719940450456407, 0.5062417984008789],\n",
       "  [-0.35338275962405735, 0.4772946834564209],\n",
       "  [-0.33587714036305744, 0.4546034336090088],\n",
       "  [-0.3052728705936008, 0.42307233810424805]],\n",
       " [[-0.23949013815985787, 0.5773220062255859],\n",
       "  [-0.23667779233720568, 0.5722372531890869],\n",
       "  [-0.22557927502526176, 0.5431935787200928],\n",
       "  [-0.20465724998050266, 0.4823343753814697],\n",
       "  [-0.1634289158715142, 0.3716559410095215]],\n",
       " [[0.1572618219587538, 0.07549667358398438],\n",
       "  [0.15792979134453666, 0.060535430908203125],\n",
       "  [0.1572618219587538, 0.0012974739074707031],\n",
       "  [0.1572618219587538, -0.09331059455871582],\n",
       "  [0.1572618219587538, -0.19061779975891113],\n",
       "  [0.16039503945244682, -0.27698779106140137],\n",
       "  [0.16449623637729222, -0.3351624011993408],\n",
       "  [0.16815900802612305, -0.36484789848327637],\n",
       "  [0.1724580923716227, -0.39047908782958984],\n",
       "  [0.17576310369703504, -0.4031686782836914],\n",
       "  [0.17781893412272134, -0.41222715377807617],\n",
       "  [0.1800888114505344, -0.4028477668762207],\n",
       "  [0.19924441973368326, -0.30954861640930176],\n",
       "  [0.23619598812527126, -0.19009900093078613],\n",
       "  [0.27544326252407497, -0.11487388610839844],\n",
       "  [0.3773776690165202, 0.006384849548339844],\n",
       "  [0.4209256172180176, 0.044335126876831055],\n",
       "  [0.515392886267768, 0.09620928764343262],\n",
       "  [0.5475585990481906, 0.1022331714630127],\n",
       "  [0.5785074498918321, 0.09987020492553711],\n",
       "  [0.6033289432525635, 0.0805196762084961],\n",
       "  [0.653514199786716, -0.06240248680114746],\n",
       "  [0.6489545769161649, -0.15761446952819824],\n",
       "  [0.6356268458896213, -0.2385087013244629],\n",
       "  [0.6132160292731391, -0.3125126361846924],\n",
       "  [0.5808015664418539, -0.3750417232513428],\n",
       "  [0.5217696560753716, -0.45111823081970215],\n",
       "  [0.48244476318359375, -0.46977734565734863],\n",
       "  [0.4216930601331923, -0.481858491897583],\n",
       "  [0.38133422533671063, -0.45923519134521484]],\n",
       " [[0.4203927516937256, 0.25031304359436035],\n",
       "  [0.423429012298584, 0.23531270027160645],\n",
       "  [0.4344024923112657, 0.17892122268676758],\n",
       "  [0.45450607935587567, 0.08088946342468262],\n",
       "  [0.4757249355316162, -0.03155040740966797],\n",
       "  [0.4615234004126655, -0.2890300750732422],\n",
       "  [0.40979835722181535, -0.5179274082183838],\n",
       "  [0.3775111834208171, -0.5808789730072021],\n",
       "  [0.3319970766703288, -0.6217114925384521]]]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "example = full_dataset[-1]\n",
    "print(data_paths._classes.iloc[example[1]])\n",
    "example[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fdeb762d-a6aa-4fb0-b8c0-b4393206bddd",
   "metadata": {},
   "source": [
    "This raw data can easily be rendered to a raster and the ordering of strokes can be conveyed with a simple hue shift where the starting point is a pure red and the hue is adjusted by a set amount of degrees within its hsv representation for each subsequent stroke."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "e473817d-3920-4a1b-8326-c12308d46117",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/jpeg": "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",
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAlgAAAJYCAIAAAAxBA+LAAAu9ElEQVR4Ae3dCXLjxpYF0KoOr9GL9CbdcPEXixKJByCRc54OR5vCkMN5KVyChL5//PB/BAgQIECAAAECBAgQIECAAAECBAgQIECAAAECBAgQIECAAAECBAgMJvDvYOM1XAIECBDoQuD/uhjF7UE8UlAW3obUAAECBJYTmCQIn3WThU8KLwgQIEDgjMDPMwd1fszH8JthYp27Gx4BAgSmEBj+jvBjCm6l2ds+RdVMggABAgSyCQwfhHsS7gj3ZGwnQIAAgVeBsYPQbd9rLb0mQIAAgQSBgYNwS8G927697QlATiFAgACBuQUGDsKtMI87wm+x9+3HuetndgQIECBwU2DUIHz9UPT19U0OpxMgQIDAagKjBuFqdTJfAgQIECgkMGQQugUstBo0S4AAgQUFhgzCoE6+IAxw7CJAgACBd4HxgtDt4HsVbSFAgACBZIHxgjCYqtvBAMcuAgQIEPgoMFgQuh38WEUbCRAgQCBZYLAgDObpdjDAsYsAAQIE9gRGCkK3g3tVtJ0AAQIEkgVGCsJgkm4HAxy7CBAgQCAQGCYI3Q4GVbSLAAECBJIFhgnCYIZuBwMcuwgQIEAgFhgjCN0OxlW0lwABAgSSBcYIwmB6bgcDHLsIECBA4FBggCB0O3hYRQcQIECAQLLAAEEYzM3tYIBjFwECBAicEeg9CN0OnqmiYwgQIEAgWaD3IAwm5nYwwLGLAAECBE4KdB2EbgdPVtFhBAgQIJAs0HUQBrNyOxjg2EWAAAEC5wVGDcLzM3QkAQIECBAIBPoNwuBzUbeDQUXtIkCAAIFLAv0G4aVpOJgAAQIECKQJjBeEbgfTKu0sAgQIEPgo0GkQBp+LfpyGjQQIECBAIE2g0yBMm4yzCBAgQIDAVYHBgtDnolcL7HgCBAgQiAV6DEKfi8Y1s5cAAQIEMgr0GIQZp6cpAgQIECAQC4wUhD4XjWtpLwECBAgkCHQXhD4XTaiiUwgQIEAgWaC7IEyeiRMJECBAgECCwDBB6HPRhOo6hQABAgQOBfoKQp+LHhbMAQQIECCQV6CvIMw7N60RIECAAIFDgTGC0Oeih4V0AAECBAikCYwRhGlzcxYBAgQIEDgU6CgIfUF4WC0HECBAgEB2gY6CcG9uPhfdk7GdAAECBO4LDBCE9yepBQIECBAgsCfQSxD6XHSvQrYTIECAQFGBXoKw6CQ1ToAAAQIE9gR6D0JfEO5VznYCBAgQyCLQRRD6XDRLLTVCgAABAgkCXQRhwridQoAAAQIEsgh0HYQ+F81SY40QIECAQCDQPgh9LhqUxy4CBAgQKC3QPghLz1D7BAgQIEAgEOg3CH0uGpTNLgIECBDIJdA4CH0umquQ2iFAgACBNIHGQZg2aGcRIECAAIFcAp0Goc9FcxVYOwQIECAQC7QMQp+LxrWxlwABAgQqCLQMwgrT0wUBAgQIEIgFegxCn4vGNbOXAAECBDIK9BiEGaenKQIECBAgEAs0C0JfEMaFsZcAAQIE6gg0C8I609MLAQIECBCIBboLQl8QxgWzlwABAgTyCnQXhHmnpzUCBAgQIBALCMLYx14CBAgQmFygTRB6UmbyZWV6BAgQGEegTRCO42OkBAgQIDC5QF9B6EmZyZeb6REgQKA/gb6CsD8fIyJAgACByQUaBKEvCCdfU6ZHgACBoQQaBOFQPgZLgAABApMLdBSEviCcfK2ZHgECBLoU6CgIu/QxKAIECBCYXEAQTl5g0yNAgACBWKB2EHpSJq6HvQQIECBQWaB2EFaenu4IECBAgEAs0EsQelImrpO9BAgQIFBIoJcgLDQ9zRIgQIAAgVhAEMY+9hIgQIDA5AJVg9CTMpOvJtMjQIDAgAJVg3BAH0MmQIAAgckFughCT8pMvspMjwABAh0LdBGEHfsYGgECBAhMLiAIJy+w6REgQIBALFAvCD0pE1fCXgIECBBoIlAvCJtMT6cECBAgQCAWaB+EnpSJK2QvAQIECBQVaB+ERaencQIECBAgEAtUCkJfEMZlsJcAAQIEWglUCsJW09MvAQIECBCIBRoHoS8I4/LYS4AAAQKlBRoHYenpaZ8AAQIECMQCgjD2sZcAAQIEJheoEYSelJl8EZkeAQIERhaoEYQj+xg7AQIECEwu0DIIPSkz+eIyPQIECIwg8FedQT4zz8ekdcD1QoAAAQInBSoFofw7WQ+HESBAgEBlgZYfjVaequ4IECBAgMC7gCB8N7Gla4F/fLzQdX0MjsB4AsU/GnXVGm9R9DpiEdhrZYyLwNgCz6dYSk1jLwiLd1xqQtqtLfAx//62gGrXQX8EphUofkc4rZyJlRf4GIHlu9UDAQJrCfiOcK16jzXb4LZPRo5VSqMl0LOAIOy5OsZGgAABAsUFBGFxYh0UEnBTWAhWswRWEygbhHtPyqymbL7JAsGno8ltOpEAAQKvAmWD8LWn19ee+HvV8DpZwE1hMp0TCRB4CrQJwmf3XhA4FHBTeEjkAAIE7ggIwjt6zm0v4KawfQ2MgMDgAoJw8AKuMXw3hWvU2SwJtBEQhG3c9ZpRwE1hRkxNEVhQoGAQemR0wfVUbspuCsvZapnA4gIFg3BP1iOjezK2JwgIyAQ0pxAg8CrQIAhfu/eawHmB18x7vva56HlARxIg8FFAEH5kmXzjv/8MP8Fv+fftx+GnZwIECFQUEIQVsfvo6pGCE2RhH5xGQYDA8AKCcPgSXprA6Pn3/ET00qwdTIAAgUCgVBB6ZDRAb7Jri8BvKfjtxyajytipT0czYmqKwFICpYJwD9Ejo3syRbfvZd7e9qKD0TgBAgS6EqgdhF1Nfp3B/Px7d67DZaFPR3draQcBAkkCgjCJzUldCvh0tMuyGBSB3gUEYe8VyjW+mW4Kc5lohwABApuAIFxoGUyThT4dXWjVmiqB8gJFgtAjo+ULl9jDNFm4N3+fju7J2E6AwJ5AkSDc68wjo3sythMgQIBAK4GqQdhqkvp9FZjjptCno6819ZoAgTsCgvCO3qjnzpGFo+obNwECnQkIws4KUms4E2ehrwlrLSL9EJhEQBBOUkjTIECAAIE0AUGY5jbDWaPfFPqacIZVaA4EOhDIH4T+dqKDsp4dwuhZuDdPn47uydhOgMC7QP4gfO/jscXfTuzJtN0eZGHbgemdAAECdQTqBWGd+eglo0D//3vcPh3NWG5NEVhWQBAuW/o/Ew9uCvvPwj/T+PrKp6NfPfxEgMCugCDcpVlqR5CFSzmYLAECCwoIwgWLfm3Knd8U+nT0WjkdTYDAm0DmIPTI6JvwMBuCm8LOs3CP2KejezK2EyDwKpA5CF+bfn3tkdFXjW5fB1nY7ZgNjAABAjcFKgXhzVE6vblAzzeFPh1tvjwMgMDQAoJw6PLlH3xwU9hzFu5B+HR0T8Z2AgSeAoLwSeHF/wQmy0J1JUCAQCwgCGMfe8cQCD4ddVM4RgmNkkA7gZxB6JHRdnXM3LObwsygmiNAoGOBnEG4N02PjO7J9Lx9uCwMbgp7djY2AgSaC9QIwuaTNIA0geGycG+aPh3dk7GdAIFNQBBaBgQIECCwtIAgXLr8h5OPbwp7+4MKn44eFtQBBAi8CwjCdxNbvggEWbgd11sWfhn6yw8+HX3B8JIAgS8C2YLQI6NfXOf6YY4snKsmZkOAQDaBbEG4NyKPjO7JzLS9n/vC4NNRN4UzLTlzIZBRoHgQZhyrphoKxDeF28D6ycKGSromQGBEAUE4YtXajHmULAxuCtvA6ZUAgb4FBGHf9elsdKNk4R6bT0f3ZGwnsLKAIFy5+ilzHz0LU+bsHAIEphYQhFOXt8zk+s/C4NNRN4VlFoVWCQwskCcI/e3EwEsgaej9Z2HStJxEgMCKAnmCcE/O307syUywvfMsDG4KJ8A3BQIEMgqUDcKMA9VUhwKdZ+GemE9H92RsJ7CmgCBcs+7ZZj1oFmabv4YIEBhfQBCOX8PWM4izMN5bdOw+HS3Kq3EC0wgIwmlK2XIie2m3t73lWH/17dPR5iUwAAL9CGQIQo+M9lPOhiN5z7z3LfWH56awvrkeCQwnkCEI9+bskdE9mVm3vybf6+s+5+umsM+6GBWB+gIFg7D+ZPTYXOCRf/2nYHMoAyBAoB8BQdhPLSYZSW8p6NPRSRaWaRAoJiAIi9FquHsBn452XyIDJFBDQBDWUNYHAQIECHQrcDcIPTLabWkN7CkQfDrqpvCp5AWBZQXuBuEenEdG92RsJ0CAAIGuBEoFYVeTNBgCwU0hHAIEFhcQhIsvANP/4dNRi4DA4gKCcPEFYPoECBBYXUAQrr4C1pl/8Omom8J1loGZEngXuBWEHhl9B7WFAAECBMYSuBWEe1P1yOiejO1tBYKbwrYD0zsBAg0FigRhw/nomkCagE9H09ycRWACAUE4QRFNgQABAgTSBQRhup0zRxTw6eiIVTNmAkUFBGFRXo2PJODT0ZGqZawE8gmkB6FHRvNVQUsECBAg0EwgPQj3huyR0T0Z2zsRCD4ddVPYSY0Mg0BNgfxBWHP0+iJAgAABAjcF0u/f9j4aTW/x5lScTuCKwOvNX3CPeKVJxxIgMKTAX0OO2qAJ5BCQfzkUtUFgeIH0+zd3hMMX3wQIECBA4McP3xFaBQQIECCwtIAgXLr8Jk+AAAECiUHoc1FLhwABAgTmEEgMwjkmbxYECBAgQEAQWgMECBAgsLSAIFy6/CZPgAABAoLQGiBAgACBpQUE4dLlN3kCBAgQEITWAAECBAgsLSAIly6/yRMgQIBAShD6I0LrhgABAgSmEUgJwmkmbyIECBAgQEAQWgMECBAgsLSAIFy6/CZPgAABAoLQGiBAgACBpQUE4dLlN3kCBAgQEITWAAECBAgsLSAIly6/yRMgQIDA5SD0R4QWDQECBAjMJHA5CGeavLkQIECAAAFBaA0QIECAwNICgnDp8ps8AQIECAhCa4AAAQIElhYQhEuX3+QJECBAQBBaAwQIECCwtIAgXLr8Jk+AAAECeYLwJ0gCBAgQIDCmwLUg3Ptr+jHnbtQECBAgQODHtSAERoAAAQIEJhMQhJMV1HQIECBA4JqAILzm5WgCBAgQmExAEE5WUNMhQIAAgWsCgvCal6MJECBAYDIBQThZQU2HAAECBK4JCMJrXo4mQIAAgckELgTh3h8R+mv6ydaE6RAgQGApgQtBuJSLyRIgQIDAIgKCcJFCmyYBAgQIfBYQhJ9dbCVAgACBRQQE4SKFNk0CBAgQ+CwgCD+72EqAAAECiwgIwkUKbZoECBAg8FlAEH52sZUAAQIEFhE4G4T+iHCRBWGaBAgQWE3gbBCu5mK+BAgQILCIgCBcpNCmSYDA7AL//jP7DEvN769SDWuXAAECBAoJyLyssO4Is3JqjAABAgRGExCEo1XMeAkQIEAgq4AgzMqpMQIECFQQ+Pn35058ZPrZ5WCrIDwAspsAAQIE5hY4FYT+iHDuRWB2BAgQWFnAU6MrV9/cCRAYWeD5AalPRO+VURDe83M2AQIEWgnIv0zypz4azdSXZggQIECgsIB0vA4sCK+bOYMAAQIEJhIQhBMV01QIEFhH4PkF4TpTLjZTQViMVsMECBAgMIKAIByhSsZIgACB8wK+Jjxv9etIQXgRzOEECBAgMJeAIJyrnmZDgMA6Ar4mzFRrQZgJUjMECBAgMKaAIByzbkZNgAABApkEBGEmSM0QIECgHwHPy1yphSC8ouVYAgQIdCXga8Ic5RCEORS1QYAAAQLDCgjCYUtn4AQIECCQQ0AQ5lDUBgECBHoT8DXh6YoIwtNUDiRAgACBGQUE4YxVNScCBNYR8LzM7VoLwtuEGiBAgACBkQWOg/Dfnen93NluMwECBAh0IeBrwnNlOA7Cc+04igABAgQIDCkgCIcsm0ETIEDgj4CvCf9YpLwShClqziFAgACBaQQE4TSlNBECBAgQSBEQhClqziFAgMAYAp6XOVEnQXgCySEECBAgMK+AIJy3tmZGgMA6Ap6XuVFrQXgDz6kECBAgML6AIBy/hmZAgAABAjcEBOENPKcSIECgfwHPyxzVSBAeCdlPgAABAlMLCMKpy2tyBAisI+B5mdRaC8JUOecRIECAwBQCgnCKMpoEAQIECKQKCMJUOecRIEBgFAHPy4SVEoQhj50ECBAgMLuAIJy9wuZHgMA6Ap6XSaq1IExicxIBAgQIzCIgCGeppHkQIECAQJKAIExicxIBAgTGEvC8zH69BOG+jT0ECBAgsICAIFygyKZIgMA6Ap6XuV5rQXjdzBkECBAgMJGAIJyomKZCgAABAtcFBOF1M2cQIEBgRAHPy+xUTRDuwNhMgAABAmsICMI16myWBAisI+B5mYu1FoQXwRxOgAABAnMJCMK56mk2BAgQIHBRQBBeBHM4AQIExhWo9rzMvz9+bP8M8n9/DTJOwyRAgACBvgXek+99y/sMfr5vqr3l4I5wbxYdjLy2lP4IECAwjECd52W2hHj9J1lnL2mSG7x4ojvCi2AOJ0CAwMoC2UPr0eD2/9vdYB3cEa5cbnMnQIAAgT8CW1ZlT8E/rRdu/LWjt9eC8I3EBgIECEwskPC8TOkIfGpX6+jZ468XgvCrh58IECBA4CnQJJmK3nc+p/byQhC+YHhJgACBaQRuPi/TJAKf+HWz0MMyT3gvCBAgQOCXwM0cOvnYS9zLtvdkO7eLJghvE2qAAAEC0wjE4bQ3zbTEepwV9PjYldb43lA/bffR6CcV2wgQIDCxwMfnZbbUCTLpo8YWUY9/Pu49ufEw566O6mS/L4e5I3zB8JIAAQILCiQkzWF6XWLcWgvGkLevTwNzR/hJxTYCBAhMIHD4vMwWP0ECfRTYYqlEMu21+dh+dZAfR76/0R3hvo09BAgQmFUgIVr2sioX0cfMSxjn9fG4I7xu5gwCBAoJVLnqFRr7zM1uEVU6BQ/5Sq4NQXjI7wACBAoIbNe1938K9KPJHYF/drZ/3dxDBH4dUYmfBGEJVW0SIEBgfIEmd4EtOhWE4y9WMyBAgMCuwN+7e4Idfd4IFvt0VBAGa8EuAgSKCey98S92sSs2k44bTsDsMwILGycGYQJv4YlongABAgR+C2zX6KuX6X4icO9N0u/JZf+3P5/ITqpBAgQItBO4mn/bSKsHT7rONrsCoz24IyzQY7qAMwkQIEBgV+DyXeA//4WKq/yPHwdBuCtuBwECBAh0InA5AjsZ9/4wgnhOuOXd7+exRxAeCdlPgACBbgVORWDSg6PdTrnAwARhAVRNEiBAoIJAgXujCqM+20VwU3i2ibPHpQfh3CU46+c4AgQINBFY+RKce+7pQdik9DolQIAAgWt/GlHx1mrQ0hwHIcNBS2vYBAjMKXD+fmi7fMdX8I//hd5+1OLB5xvncRDm60tLBAgQIHBP4GQKHkbgvVG0P/ukw7mBCsJzTo4iQIBAc4EzV/+PEXj4X+htPrWmA7gVhGeK0nR2OidAgMAsAocX3I8ROPrsq3w6eisIRxc2fgIECIwhcCYFx5hJvlEempzu6lQQVonk00N2IAECBJYSOLzi37lGd/68TJVCnwrCKiPRCQECBI6eclxNaIvAoik4hOedmD83QUF4zslRBAjkFTi8vuftbsTWzhCdD4kpn5c5Q3Si9ILwBJJDCBAgUFngzCX+fApWHvxo3d0NwjPFGs3EeAkQINBU4MyFdakULDzZu0HYdLHonAABAtMJHKbglgqFg2Ek00OuE5M5G4TYT2A6hAABAvcEDi/rJa7Fyz84ejYI79XW2QQIECBwJNAkBY8G1cv+Eu8Afs9NEP6W8G8CBAg0FKiQglM+OLqV7JDuqKyC8EjIfgIECJQWOLyUl7wfKj25/tvPEISHFexfwQgJECDQTODwGioFH7UJHA4Nw+pmCMKwfTsJECBAYF/g8AoeXP33W728Z+3nZS4EYZ1yXK6fEwgQIDCrgMvut8qWAbkQhN/G40cCBAgQuCUQ3w6WuOjP+rzMVoYYM6yTIAx57CRAgEAhgfjCXSIFC01k/GbzBGFc0PGVzIAAAQJZBeKLphQMsAvg5AnCYMx2ESBAgMAXASn4hSPfDzHsfj/XgrBAEu8PzR4CBAjMJ5B6sS4usfCDo9eCsHgldECAAIGJBQ5T0N3GmernVhKEZ9QdQ4AAgfICua/vn0c88YOj24QP32p8QhGEn1RsI0CAQHaB+BpdJwWzT2qKBrMFYVziKaxMggABAqkC8SVSCl51zSqWLQivzsLxBAgQWEVACtasdKz9aSSXgzBrDH8akW0ECBCYSSC+Lnd1SV31wdHLQTjT+jQXAgQIlBWIU7Bs37O3nu89hCCcfa2YHwECrQQOUzDfpfzaFOd+cHSzOJT/6pUzCC92/XUgfiJAgMBSAq1ScCnkc5PNGYTnenQUAQIEFhCI7wykYJYlkIkxJQgzdZ2FQSMECBDoT0AKNq9JXIKvw0sJwq8t+IkAAQIEXgTiS7A7iReqTl4Kwk4KYRgECEwhIAUrlzF4YxHX4mWcgvAFw0sCBAjcETh95b3TiXOzC2QOQssge4U0SIDAJALBvcskM2w0jduwmYOwEYNuCRAg0Fogvg+4fbFuPb0x+4+L8ntOiUGopr8B/ZsAAQJHf8Htitn3GkkMwr4nZXQECBCoKBDfdnSYgnv/m6Lj/i/O3EMWhBV/W3RFgMB8AsOl4HwliGcUF+jXuYIwJrSXAAEC+wLxRfbebcp+r/ZkFsgfhPHCyDx8zREgQKCVgItdK/mP/d5425EehDc6/TgJGwkQIDCRgEvkOMVMD8Jx5mikBAgQyC0Q3w5KwdzeRdsThEV5NU6AwIwCUnCuqgrCueppNgQIlBaQgqWFq7dfJAjjdVJ9jjokQIBAJoH46uYT0UzMlZspEoSV56A7AgQI1BCIU7DGCPRRROBWEHr3U6QmGiVAoEOBwxR0QeywaueGdCsIz3XhKAIECMwuIAVHrrAgHLl6xk6AQB2B+HZQCtapQrFeBGExWg0TIDCHgBSco477sygVhPHK2R+PPQQIEOhJIL6WuRfsqVbJYykVhMkDciIBAgR6EZCCvVSi7DjuBqH3Q2Xro3UCBFoJxCnYalT6LSBwNwgLDEmTBAgQaC1wmILj3gTM91/lvb1YBOFtQg0QILCawLgpuFqlzs23YBAevqM6N0JHESBAoK5AfPGSgnWrUaG3gkFYYfS6IECAQGYBKZgZdIDmMgSht0cD1NkQCRA4IyAFzyhNd0yGIJzOxIQIEFhSYEvB4H19sGtJrWEmfaJwgnCYahooAQIFBR73gvEdYcHuNX1b4EbtBOFtfQ0QIDC9wIm7iukNJp5g2SC8kdATm5saAQKdCcSXKinYWbmyD6dsEGYfrgYJECBQVUAKVuVu01meILRU2lRPrwQI3BeIbwfvt6+F7gXyBGH30zRAAgQIXBfwHv+62YhnCMIRq2bMBAhkEnA7mAly6GaKB6FlNvT6MHgC6wq4HVym9sWDcBlJEyVAYDQB79NHq1ih8WYLQm+eClVIswQINBBwRWuA3qzLbEHYbAY6JkCAQIKA28EEtElPEYSTFta0CBBIFpj4dtB/lffTqhCEn1RsI0BgbgG3g3PX9+LsagShJXexKA4nQKCdwMS3g+1QO++5RhB2TmB4BAisJeC9+Vr1Pp5tziD0RurY2xEECLQViFPQVaxtdRr1njMIG01BtwQIECBAIF1AEKbbOZMAgcEE3A4OVrBKwxWElaB1Q4BA1wI+FO26PGUHVykI4/dhZaeodQIECGwCLkOWwY5ApSDc6d1mAgQIdCDgdrCDItwawr13OZmD0HK6VUsnEyBQSODehbLQoDTbiUDmIOxkVoZBgACBswLev5+VGvC4c8UVhAOW1pAJELgk4HbwEtd6B9cLQktxvdVlxgS6Fzh3x9D9NAzwlkC9ILw1TCcTIEAgTcB78DS3lc7KH4TeYK20fsyVwMgCrlYjV+/P2G+/18kfhH8G5xUBAgTaCty+RLYdvt7rCAjCOs56IUCgMwG3g50VJP9wTpdYEObH1yIBAl0IuB38Vgb/efpvIL9/rBqEluVvdv9uIGD5NUBv3uXePcHe9uYDNoAWAlWDsMUE9UngP4FHCm7/XxyutSCe9ZZ8sxb+WeIbEywShJbcjYo4tbjAIw5z/PoUH6oO0gW+Ffj1R1eodNY5zywShAHV62oMDrOLQEaBvVW3bd/blbF3TREg0EDgytud2kHYgEOXBPYFrvyy7Ldiz0ACSj5QsWoNtUEQeg9eq7j6+U/AeltxHaj6ilVPn3OpIPSuK70mzqwlYJXWktYPga4FSgVhPGlv12Ife3MJWGm5JCdpx3ufSQr5exqZfsMLBqEl97tW/t2jgPXZY1WMiUALgYJBGE8nU5DHndi7tIA1tmj5FX7Rwr9M++L73LJBeHEwL9PwkkBJASuzpG6vbat6r5VpPq6yQRhPz/u22MfeOwJW1x29GufuxZLK1dDXxxeB4kG4t9q/jMIPBCoKWJMVsat3JUerkzfrMF+tiwdhbJRvInE/9q4lYF2tVW+zJXBPoEYQegN+r0bOzilgNebUHKgthR+oWDeHer3WNYIwnpQ377GPvVcFrKirYvMcr/bz1LLqTCoF4fWErqqgs0UErMNFCm2aHwT8V3k/oPxvU6Ug3B/Af3u8jYt97D0vYC2dt1roSO+AFip2ylTrBaGlmFIf51wRiFPQCrxiOeCxcfkHnJAhRwJZy10vCKMpuSmMdewlQIAAgWICVYPQW/JiddTwwQfs1t66S0Ttl6p9UrmrBmFcjqx3unFX9q4lkPSrsRbR8LN1+Ri+hC0nUDsIXZJaVnvevl0G562tmRF4E8j9C187CN8m9GVD7tl9adwPawp477Vm3f83a+VfuvxnJ98gCK3Ms8Vx3DkB75/OOTmKwOwCqenSIAjjUrioxT72XhJI/b241ImDWwu4arSuQNX+C5S7TRC6PFVdN1N3VuCXYmqvpSbnQrNUuW9Mtk0QxgN2aYt97D0p4DJ4EsphBGYQuPEL3ywIb4x5hpKZQxYB75myMDZrZO8qcKmulw5uNlUddy3QLAhjFWs79rH3UGDvGnt4ogMIEOhXoEw2tAzC10vV6+t+a2BkPQmU+Y3oaYbGckfANeWO3mLntgzCB/W2XLd/3i9q71sWK43ppgu4BqbbjXWmy8RY9So32nu/842D8BGBe4t5b3s5TC2PImBtjFIp4ySQTaDYr33jINyA7gV5NmENTSNgRU1TyvSJWATf7PxXeb+BfP2xfRB+Hc/3n4q9A/jekZ8HErAqBipWwaFaBwVxh2r69vueLoLw9iyGqpnB3hOIr37W0j1dZxPoVSD+zb836i6CcJtCcP0qOf17eM7uTCBYRZ2N1HBKClgHJXWnbLuXIIxxZWHss85eK2GdWh/M1FI4AFpmd473PR0FYY7pLFN7E30TsH7eSGwgMItA4fc9HQVhXLHCDnHn9nYhYA10UYbOB+ENUecF6nJ4fQVhvIZdB7tcQl0MKl45XQzRIAgQyC6Q6Te/ryDclDLNK7u3BhsLeBvUuABddW81dFWO0oMpX+7ugjAmLQ8S929vjwLePPVYlSZjshSasI/faY9BaDGPv64yz8AboMygmltKYO9/VmZ0hHxR0WMQbtUJJuiaOPrqzTv+YKnk7UhrvQi4BPRSiXnG0WkQxsB+EWKfyfYq92QFNZ1eBH7+3ctIgnFU+f3vNwi90w/Whl0PAYvESvgjYDX8sfDqmkC/QRjPo8q7hHgI9tYQUOgaygP1YUFcLZYvCE+IdR2E8Ts8vxEn6jvzIfHymHnmM81NFWeqZva51LrKdx2Em2r8a1JLKXt5NXhKQH1PMU150NXax1eKKYluTmqILwhvzvH06b0H4eFErv6+HDbogE4E4sq67nVSJsMgUEPg2y/8tx9vj2CAIMw95dtmGiBAoIlA/OaoyZB0Wk7gtdzP12XyYIAg3JzjuT+JylVEy5UF4prG66HyUHVHoF+B+Z6UiS8NqZUYIwi32cXXvjI4qajOKykQr4SSPWu7VwFrotfKpI8ruKYXKPcwQbiBxtMP3NKL4cwWAkrZQl2fKwl4UuZrtUcKwm3ksvBr+Zb7KV4Ay3GYMIEpBYL3wmUuAYMF4WHRA8DDcx3Qg4AK9lAFYyCwlMB4QXj4hsCVdNYVfFj6WSduXv8J+MW+ug4GfVImKHSxS8B4QbgthkONQPLqWnJ8TQGFq6mtLwIEHgJDBuE29MMsVODJBFR8soJmm46VcZWy5ydlgvfCJQs9ahBupY9ZAs+ry8bxdQSUrI6zXggQ+CYwcBBuM5GF38o5649xoWedtXkRSBcY8QvC4L1w4UvA2EG4rZLYJ4BNX2HOLCCgUgVQB2/ydU28vh58WobfocDwQbiZysIOF1bGIcX1zdiRpgjML9DtF4TBe53yl4AZgvBw7QbCh+c6oIKAAlVA7reLO5e5O+f2K2JktQUmCcLDXweX2torK1N/h5XN1I9mCBBoJxBcoKtcAiYJwq2Ah1wbdaDdbgms3rOirL4CzD+7wIhPymRHuNLgPEG4zfowC7djXHavLI/Gx54paOMh6p4AgZsCwUW51iVgqiDcynHGLWC/WVCn3xQ4U76bXTh9JIHH76rf2Fw16/ZJmVwTTG1ntiDcHLaL6eH11G9W6oLJfN63Qjx+fJTvsIiZh6K54QQskeFK9j7gb5eA1wMq1nfCIHxIboYx4+YflOC1HF4XEtjz37bHtSs0Hs02E1DvjPS+ILyOOW0QPigOf7/2rsXXJZ1xTSCQP6zatZ4cTYBAnwLdXAUmD8Kt+odX1aAWfS6eCUbFfIIimsJgAr4g3C/Y/EG4zV0W7i+A2nu2CIxT8LBYtUesv7YC8XJpOza93xEIKlv9KrBEEG7F2mBj28ML9J2KO/chEKz8xwFxjTASIECghMAqQfiwO7zOHl6pS9RghTbPvM84rM4KUOvO8Wr5rx6/iOwoT8oEl9oWlV0rCLffhUPkoECL/CrlneaZCMzbo9YIEOhaoL+L7HJBuK0PWVjnl+RkBD7KcViUOmPWC4E5BUZ5UqbRheCvOat+NKuHdvC+5LGrUVGORt/9/gD2fezbwZzfWWwhMKfApatDLYJFg/DBu11/46I89rpMn1mNseReC2z3ZGwnMKFAcJloei1YOgi3dXaYhdsxz9o1rVR3vxRPluSR8UymW/1ES+fjCuj8SZngktG6oKsH4bacthIEBXpdb4/DWpfsdUS1X5+EOhzWyoaHOOsecP5XcV0jMy8iIAj/Y31cl09e5Z+HrXM1f075/hpcB+2+lRYI5BHo4UmZ4CLSwUVBEP5ZaVs5gmL9Oe73q8fBHRTx94By//uSxmHnE0Mdzt0BBJYWCC4lfVwXBOGX9Xk1C7eT54vDYNF+wbryQx+r/cqIHUtgLIFuvyAMLijdXBcEYZ7F/qx1N5W9Nq/n+K+dduLoQUFOzMwhBAhMIiAIvxfyceFODobHiaNc/ZOn+V3t5edR5v4yZC+7EdhWz+GitMIulavtF4RBNXuqoyD8vKaeNQrq+PnMX1ufZz3bCQ6uv+s5vCxd9znHLFPTCAEC6QLBhaazq4YgPKjyo15BQePzHyc2LHryyON5NZxRPDB7CRDoQiC49PR3+RCEp9bMs3BBcYOGzpz17CJoZ2/Xmfb3zr20/c4gL3XkYAIELgh0+6TMhTm0PFQQXtN/JEGJ4CnR5rW57R8t//Zt7MktsK2211+G9x9zd6i9/AKvFfzWepdXE0H4rUqnfnyUMqj1qVb6PqjL5do3mdFlFHhdf8/Xc//KZdR7NtXkSZmgTM9SPkfYxwtBmF6HZ02Duqe33ujM56Qa9a9bAr//p57eJazOd5PetgRXw47LJwgzrKNHfYMFkKGPwk10vEQLz1zzBEYX6OcLwuAi2PclRhBm+yV4Fvq5GLYtz9fZusna0HPMWVvVGAECBEYSEIT5q/VIl34iUNrlr7EWCfQvUPkLwuCS1/01SBCWWs6P0p9cAMESOjm+kx2dbM1hBAgQuCAQXMJGuDYJwgu1LnfoyaWyLbaTR5YbqpYJECDwRWDwFNzm8n9f5uOHvgWkYN/1MToC1QX6eVKm+tQzdigIM2JqigABAosJjH87uBVMEC62ak2XAIEVBOo8KTNFCm7LQRCu8DthjgQIEMgtMEsKbi6CMPfi0B4BAgQIDCUgCIcql8ESIEDgKdDwSZmJbgc3Tn8+8VxTXhAgQIDAkUAQgdupYz7aLgiPqm4/AQIExhIo9KRMHIEb0ZgpuA3cR6NjLXCjJUBgDYHD1KnJsA2mq/Hknrs7wtyi2iNAgEAWgUf27N1m1fmC8Hz+7Y0zC0XhRtwRFgbWPAECBK4KvMbP9vr1x6tNJR9/st9H/o2cgpuQIExeJk4kQIBALYHzWXj/C8KTEfiY+nbw4Cm4zcNHo7XWsX4IECBwRmAv8x7bi6bOXtdnhj3yMYJw5OoZOwECqwkUisO0CCyayhUrKwgrYuuKAAECscDJQLr/pMzJjvZGO0sEPuYnCPfqbDsBAgSqCzwC5mZKvY86Y4NzReCDShC+LxlbCBAg0FQgOQ63J2UyZt43gxkj8DFFT41+K7UfCRAg0IdAQvAUSsFtJAmD6UPxzCjcEZ5RcgwBAgRaCDzi53u8/VNvKFPn35NRED4pvCBAgECXAp/jsPBQ14jAB6IgLLyYNE+AAIEsAsdx+HeWfub+FPQjkSD8yGIjAQIEuhTY4vD7J6U3xrnSbV/AJAgDHLsIECAwi4DM26+kINy3sYcAAQK9CZz8U3qxd6VwgvCKlmMJECDQrYDwSy2NvyNMlXMeAQIE+hG4/x+d6Gcu1UciCKuT65AAAQIEehIQhD1Vw1gIECBAoLqAIKxOrkMCBAikCZx8Uiat8YXPEoQLF9/UCRAgQODHD0FoFRAgQGBwAU/K3CugILzn52wCBAgQGFxAEA5eQMMnQGARAV8QFiu0ICxGq2ECBAgQGEFAEI5QJWMkQIDAnoAvCPdkTm8XhKepHEiAAIFWAq+fi0q+3FUQhLlFtUeAAIGiAs9QlIiZnAVhJkjNECBAoJDAM/m+tb+3/dthfjwSEIRHQvYTIECgWwE3hTlKIwhzKGqDAAEChQTc9hWCfWlWEL5geEmAAIGBBNwOZiqWIMwEqRkCBAhkF3A7mJ30U4OC8JOKbQQIEOhcwO1gvgIJwnyWWiJAgEBGAbeDGTHDpgRhyGMnAQIEOhRwO5i1KIIwK6fGCBAgkEXA7WAWxnONCMJzTo4iQIBAJwJuB3MXQhDmFtUeAQIEbgq4HbwJePF0QXgRzOEECBBoKOB2sAC+ICyAqkkCBAgkC7gdTKZLPVEQpso5jwABApUF3A6WAReEZVy1SoAAgQQBt4MJaLdPEYS3CTVAgACBCgJuB4shC8JitBomQIDAJQG3g5e48h0sCPNZaokAAQKFBNwOFoL91awgLKmrbQIECJwUcDt4EqrAYYKwAKomCRAgkFHA7WBGzE9NCcJPKrYRIECgpoDbwZrab30JwjcSGwgQINCPgNvB8rUQhOWN9UCAAIFAwO1ggFNllyCswqwTAgQIfBSIU9Dt4Ee03BsFYW5R7REgQOCkQJyCJxtx2G0BQXibUAMECBBIEAhS8HEj6HYwQTXplJ9JZzmJAAECBFIFggh8NikFnxTlX7gjLG+sBwIECDwFpOCTopsXgrCbUhgIAQLTC0jBLkssCLssi0ERIDCfgBTstaa+I+y1MsZFgMBMAocp6EvBduV2R9jOXs8ECCwiIAX7LrQg7Ls+RkeAwOgCUrD7CgrC7ktkgAQIjCsgBUeone8IR6iSMRIgMJzAYQRuM/K9YB9ldUfYRx2MggCBmQSk4FDVFIRDlctgCRDoX0AK9l+jryMUhF89/ESAAIE7AlLwjl6jc31H2AhetwQIzCoQZKEvBbssujvCLstiUAQIjCuwl3Z728ed6SwjF4SzVNI8CBDoR+A989639DPa5UciCJdfAgAIECgh8Jp8r69L9KVNAgQIECDQqUDwfWGnIzYsAgQIECBAgAABAgQIECBAgAABAgQIECBAgAABAgQIECBAgAABAgQmFvh/7AFVUQBX0sAAAAAASUVORK5CYII=",
      "text/plain": [
       "<PIL.Image.Image image mode=RGB size=600x600>"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "renderer = StrokesToPil((600, 600), stroke_width=8, invert_y=True, multicolor=True, mc_hue_shift=40)\n",
    "renderer(example[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d90bac51-ee0f-4b90-8098-0c1a3911cac0",
   "metadata": {},
   "source": [
    "While the ragged datastructure of strokes with varying counts of points can be encoded easily in a raster image, using the raw coordinates is ill suited for the consistently shaped tensors necessary for an RNN. It's possible to pad the total amount of points to a consistent value, but there is no defined upper limit. The amount of points captured depends entirely on how long it takes to draw the stroke.\n",
    "\n",
    "Because more informative features such as size and relative angles can be extracted from the coordinates alone, the temporal aspect in conjunction can be viewed almost as a form of noise that says more about the person drawing than the actual character itself. This could also increase the possibility of overfitting during training.\n",
    "\n",
    "In order to simultaneously eliminate the need for empty points while also eliminating the temporal information encoded in the data, a transformation is applied for the RNN models where each individual stroke is represented by a uniformly lengthed b-spline interpolated from the sequence of x and y points. The cell below displays a visualization of the same datapoint, but this time using its b-spline representation. The two images are almost identical with the exception of a slight smoothing introduced from the addition of new points along the fitted curve (which in many cases may actually be a better representation of the drawer's intent)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "a2260f15-e394-4464-b1ce-6a3594bf5f01",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5, 120, 2)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[-0.63213074,  0.83440471],\n",
       "       [-0.63319596,  0.83321866],\n",
       "       [-0.63548936,  0.82979721],\n",
       "       [-0.63881829,  0.82434482],\n",
       "       [-0.64299011,  0.81706598],\n",
       "       [-0.64781219,  0.80816515],\n",
       "       [-0.65309186,  0.79784683],\n",
       "       [-0.65863649,  0.78631549],\n",
       "       [-0.66425343,  0.77377559],\n",
       "       [-0.66975656,  0.76043016]])"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "b_spline_transformer = ToBSplines(120, 3, smooth_points=.5)\n",
    "\n",
    "# Located in the cleaning functions python file, this function fixes the orientation and converts each raw points dictionary\n",
    "# into a list of their interpolations along a b-spline.\n",
    "example_b_spline = b_spline_transformer(example[0])\n",
    "\n",
    "#The shaoe is consistent and the coordinates should be sufficient for each observation.\n",
    "print(example_b_spline.shape)\n",
    "    \n",
    "#Looking at the first few datapoints in this notably long and complicated stroke, you can see that there are now more datapoints.\n",
    "example_b_spline[0][0:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "b721edce-c472-4fe4-b457-f0fe552a605c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/jpeg": "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",
      "image/png": "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",
      "text/plain": [
       "<PIL.Image.Image image mode=RGB size=600x600>"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "renderer(example_b_spline)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6fa0c89b-6206-41ba-a74a-3ec645f2b222",
   "metadata": {},
   "source": [
    "## Baseline: Dummy Classifier"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5e328a96-1866-4436-b630-2d265d756906",
   "metadata": {},
   "source": [
    "A dummy classifier that predicts the dominant class is used to establish a prediction baseline."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "57e05d43-c384-4771-9a36-d5c22b189e98",
   "metadata": {},
   "outputs": [],
   "source": [
    "def balanced_acc_baseline(X, y, strategy=\"prior\", random_state=None):\n",
    "    baseline = DummyClassifier(strategy=strategy, random_state=random_state)\n",
    "    baseline.fit(X, y)\n",
    "    y_pred = baseline.predict(X)\n",
    "    acc = accuracy_score(y, y_pred)\n",
    "    balanced_acc = balanced_accuracy_score(y, y_pred)\n",
    "    print(\"Accuracy:\", acc)\n",
    "    print(\"Balanced Accuracy:\", balanced_acc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "3d4a0a02-e2ca-4699-90d8-0b85d6874fc3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.012060750446694461\n",
      "Balanced Accuracy: 0.009259259259259259\n"
     ]
    }
   ],
   "source": [
    "X = [item for item in full_dataset]\n",
    "y = full_dataset.labels\n",
    "balanced_acc_baseline(X, y, random_state=SEED)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a09fdbfa-8678-40a5-b6b8-e6c7139481e5",
   "metadata": {},
   "source": [
    "## Model 1: CNN (VGG16)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7ce23bf9-fc12-4d09-b52d-d9413c2671e9",
   "metadata": {},
   "source": [
    "The early prototype of the character identifier CNN explored the use of leveraging an existing hiragana/kanji dataset to facilitate transfer learning on the much smaller custom dataset. While in many cases this would be a great practice, this required the loss of features present in the custom data. By using an app to capture hand drawn input there is definitive information about stroke order and direction that would not be present in handwritten scans. The following model explores integrating stroke order through the use of hue shifted color channels.\n",
    "\n",
    "To optimize training, transfer learning is directly applied using VGG16 weights trained on ImageNet. Because the training images are drastically different from the typical ImageNet classification objectives, the classifier layer is dropped entirely in favor of a more simplistic replacement and only the two lowest level convolutional layer groups are frozen."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "9e0a18ac-807e-4831-abc7-7650e2f34b19",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torchvision.models as models\n",
    "\n",
    "cnn = models.vgg16(weights=\"VGG16_Weights.IMAGENET1K_V1\")\n",
    "# Freeze a;; layers in the CNN\n",
    "for param in cnn.parameters():\n",
    "    param.requires_grad = False\n",
    "# Unfreeze the third block onwards for training.\n",
    "for i in range(10, len(cnn.features)):\n",
    "    for param in cnn.features[i].parameters():\n",
    "        param.requires_grad = True\n",
    "\n",
    "class ImageAggregator(StrokeDatasetAggregator):\n",
    "    def __init__(self,\n",
    "                 data_paths,\n",
    "                 image_width,\n",
    "                 stroke_width,\n",
    "                 multicolor,\n",
    "                 mc_hue_shift=40):\n",
    "        trans =  transforms.Compose([StrokesToPil((image_width, image_width), stroke_width=stroke_width, invert_y=True, multicolor=multicolor, mc_hue_shift=mc_hue_shift), transforms.PILToTensor(), transforms.ToDtype(dtype=torch.float32, scale=True)])\n",
    "        super().__init__(data_paths, trans, None)\n",
    "        if multicolor:\n",
    "            self._multicolor_reshape(image_width)\n",
    "\n",
    "    def _multicolor_reshape(self, image_width):\n",
    "        self._data = self._data.reshape(-1, 3, image_width, image_width).permute(0, 2, 3, 1).reshape(-1, 3)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fddcb1e8-3e19-4b5f-9b46-14995a60f283",
   "metadata": {},
   "source": [
    "To prepare the data with these weights, the inputs need to be normalized. The following cell extracts the rgb means and standard deviations from the entire dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "5ac87b38-1f4e-4385-9700-69d2264d974f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean: tensor([0.0390, 0.0239, 0.0026])\n",
      "STD: tensor([0.1890, 0.1414, 0.0468])\n"
     ]
    }
   ],
   "source": [
    "aggregator = ImageAggregator(data_paths, 224, 4, multicolor=True)\n",
    "mean_rgb, std_rgb = aggregator.get_mean_std(dim=0)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a2e90532-7053-4103-87e5-8b1e7e676971",
   "metadata": {},
   "source": [
    "But because each stroke should be weighed equally, and because the hue rotation ensures that the red channel will be much more common than the other two, I opted to normalize the data using the mean and standard deviation across all color channels instead of each individually. Otherwise the presence of the blue channel appearing in later strokes would be a stronger input signal by a factor of 4."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "c46963b2-401c-4435-ae8b-994b732a75b7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean: tensor(0.0489)\n",
      "STD: tensor(0.2157)\n"
     ]
    }
   ],
   "source": [
    "aggregator = ImageAggregator(data_paths, 224, 4, multicolor=False)\n",
    "mean_gray, std_gray = aggregator.get_mean_std()\n",
    "norm = transforms.Normalize(mean=mean_gray.unsqueeze(-1).unsqueeze(-1), std=std_gray.unsqueeze(-1).unsqueeze(-1))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f3a09c1f-ec26-43be-b641-39683cdd658b",
   "metadata": {},
   "source": [
    "Next, custom dataset and dataloader dictionaries are created with data augmentation occuring within only the training set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "9219e288-cf8f-4309-a80f-2e3984e794bb",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'train': <torch.utils.data.dataloader.DataLoader at 0x7ffa132bed90>,\n",
       " 'val': <torch.utils.data.dataloader.DataLoader at 0x7ffa132c9890>,\n",
       " 'test': <torch.utils.data.dataloader.DataLoader at 0x7ffa1bff6f10>}"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# These transforms will be the common start to the train, val and test sets.\n",
    "test_transforms = [StrokesToPil((32, 32), stroke_width=2, invert_y=True, multicolor=True, mc_hue_shift=40),\n",
    "                   transforms.PILToTensor(),\n",
    "                   transforms.ToDtype(dtype=torch.float32, scale=True),]\n",
    "# Training dataset will be augmented by random transformations.\n",
    "train_transforms = transforms.Compose(test_transforms + [transforms.RandomAffine(degrees=4, translate=(.1, .1), scale=(.9, 1.1), shear=.3), norm])\n",
    "test_transforms = transforms.Compose(test_transforms + [norm])\n",
    "transformed_dataset = StrokeDataset(data_paths.get_data(), transform=test_transforms)\n",
    "split_keys = {\"train\": train_transforms, \"val\": test_transforms, \"test\": test_transforms}\n",
    "# The dataset has a 70% train, 10% validation, and 20% test split.\n",
    "split_datapaths = dict(zip(split_keys, data_paths.train_val_test_split(train_size=.7, test_size=.2, random_state=SEED)))\n",
    "datasets = {key: StrokeDataset(split_datapaths[key], transform=value) for key, value in split_keys.items()}\n",
    "dataloaders = {key: torch.utils.data.DataLoader(datasets[key], batch_size=batch_size, shuffle=True, num_workers=5, drop_last=True) for key in split_keys}\n",
    "dataloaders"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "75a56b5c-a02a-4aae-84df-56d52b0de99c",
   "metadata": {},
   "source": [
    "The new classification layer consists of two Dense layers. The output of the first layer has a Relu activation function and a dropout layer for regularization. Finally like all of the subsequent models it uses the Softmax function to choose between the available classes. Also common to each model is cross entropy for the loss function, Adam for the optimizer, and a step scheduler to decay the learning rate."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "cbefa76e-7363-4676-8df7-e5beb4023337",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "108\n"
     ]
    }
   ],
   "source": [
    "# The input size of the previous classification layer group.\n",
    "n_features = cnn.classifier[0].in_features\n",
    "# Amount of classes to predict.\n",
    "n_classes = np.unique(full_dataset.labels).size\n",
    "print(n_classes)\n",
    "dense_hidden = 40\n",
    "learning_rate=0.00006\n",
    "cnn.classifier = nn.Sequential(nn.Linear(n_features, dense_hidden), nn.ReLU(inplace=True), nn.Dropout(p=0.5, inplace=False), nn.Linear(dense_hidden, n_classes))\n",
    "#cnn.classifier[-1] = nn.Linear(n_features, n_classes)\n",
    "criterion = nn.CrossEntropyLoss()\n",
    "optimizer = torch.optim.Adam(cnn.parameters(), lr=learning_rate)\n",
    "scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=11, gamma=0.4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "a5e6f0e8-45c4-42d5-80fa-d2b87210f7bf",
   "metadata": {},
   "outputs": [],
   "source": [
    "trainer = CharacterTrainer(cnn, optimizer, criterion, dataloaders, scheduler=scheduler, single_input=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "554ce55f-6ebd-4fbd-b98a-46ce4d9b93ca",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "1        0.199804   0.102612   5372.0   0.11455  0.472727  448.0  0.00006\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "2        0.117295   0.388246   5372.0  0.039384  0.843182  448.0  0.00006\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "3        0.066997   0.625373   5372.0  0.017735  0.934091  448.0  0.00006\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "4        0.051026   0.708582   5372.0  0.012627  0.945455  448.0  0.00006\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val       lr\n",
      "epoch                                                                   \n",
      "5        0.042064   0.759701   5372.0  0.007379    0.975  448.0  0.00006\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "6        0.034712    0.79347   5372.0  0.007104  0.965909  448.0  0.00006\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "7        0.030932   0.815485   5372.0  0.002055  0.979545  448.0  0.00006\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "8        0.025177   0.846642   5372.0  0.001502  0.993182  448.0  0.00006\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "9        0.025183   0.850187   5372.0   0.00117  0.995455  448.0  0.00006\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "10       0.025113   0.848694   5372.0  0.001879  0.995455  448.0  0.00006\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "11       0.023708   0.860075   5372.0  0.002284  0.984091  448.0  0.00006\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val        lr\n",
      "epoch                                                                    \n",
      "12       0.018856   0.876866   5372.0  0.000016      1.0  448.0  0.000024\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val        lr\n",
      "epoch                                                                    \n",
      "13       0.016996   0.888993   5372.0   0.00002      1.0  448.0  0.000024\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val        lr\n",
      "epoch                                                                    \n",
      "14       0.015972   0.898134   5372.0  0.000005      1.0  448.0  0.000024\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val        lr\n",
      "epoch                                                                    \n",
      "15       0.016419   0.896269   5372.0  0.000007      1.0  448.0  0.000024\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val        lr\n",
      "epoch                                                                    \n",
      "16       0.014811   0.909701   5372.0  0.000044      1.0  448.0  0.000024\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val        lr\n",
      "epoch                                                                     \n",
      "17       0.016481   0.897015   5372.0  0.000637  0.997727  448.0  0.000024\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val        lr\n",
      "epoch                                                                     \n",
      "18       0.015402   0.901493   5372.0  0.000103  0.997727  448.0  0.000024\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val        lr\n",
      "epoch                                                                    \n",
      "19       0.013993   0.907276   5372.0  0.000013      1.0  448.0  0.000024\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val        lr\n",
      "epoch                                                                    \n",
      "20       0.013496   0.911194   5372.0  0.000006      1.0  448.0  0.000024\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val        lr\n",
      "epoch                                                                     \n",
      "21         0.0136   0.911381   5372.0  0.000795  0.997727  448.0  0.000024\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val        lr\n",
      "epoch                                                                    \n",
      "22       0.013086   0.914179   5372.0  0.000009      1.0  448.0  0.000024\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val       lr\n",
      "epoch                                                                   \n",
      "23       0.012593   0.915299   5372.0  0.000052      1.0  448.0  0.00001\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "24       0.014211   0.910261   5372.0  0.000258  0.997727  448.0  0.00001\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "25       0.011789   0.919403   5372.0  0.000238  0.997727  448.0  0.00001\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "26        0.01166   0.921269   5372.0  0.000131  0.997727  448.0  0.00001\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "27       0.012474   0.914179   5372.0  0.000129  0.997727  448.0  0.00001\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "28       0.012148   0.920522   5372.0  0.000151  0.997727  448.0  0.00001\n",
      "       train_loss  train_acc  n_train      val_loss  val_acc  n_val       lr\n",
      "epoch                                                                       \n",
      "29       0.010933   0.926679   5372.0  8.473287e-07      1.0  448.0  0.00001\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val       lr\n",
      "epoch                                                                   \n",
      "30       0.011522   0.923507   5372.0  0.000012      1.0  448.0  0.00001\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "31       0.011791   0.925187   5372.0  0.000236  0.997727  448.0  0.00001\n",
      "       train_loss  train_acc  n_train      val_loss  val_acc  n_val       lr\n",
      "epoch                                                                       \n",
      "32       0.011458   0.922015   5372.0  5.158910e-07      1.0  448.0  0.00001\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val       lr\n",
      "epoch                                                                   \n",
      "33       0.011378   0.922015   5372.0  0.000086      1.0  448.0  0.00001\n",
      "       train_loss  train_acc  n_train      val_loss  val_acc  n_val        lr\n",
      "epoch                                                                        \n",
      "34       0.010526   0.925746   5372.0  2.282475e-07      1.0  448.0  0.000004\n",
      "       train_loss  train_acc  n_train      val_loss  val_acc  n_val        lr\n",
      "epoch                                                                        \n",
      "35       0.010228   0.931343   5372.0  4.299960e-07      1.0  448.0  0.000004\n",
      "       train_loss  train_acc  n_train      val_loss  val_acc  n_val        lr\n",
      "epoch                                                                        \n",
      "36       0.011143   0.922201   5372.0  2.418208e-07      1.0  448.0  0.000004\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val        lr\n",
      "epoch                                                                    \n",
      "37       0.010649   0.926866   5372.0  0.000001      1.0  448.0  0.000004\n",
      "       train_loss  train_acc  n_train      val_loss  val_acc  n_val        lr\n",
      "epoch                                                                        \n",
      "38       0.010454   0.928918   5372.0  1.563356e-07      1.0  448.0  0.000004\n",
      "       train_loss  train_acc  n_train      val_loss  val_acc  n_val        lr\n",
      "epoch                                                                        \n",
      "39       0.010814   0.925746   5372.0  2.169874e-07      1.0  448.0  0.000004\n",
      "       train_loss  train_acc  n_train      val_loss  val_acc  n_val        lr\n",
      "epoch                                                                        \n",
      "40       0.010376   0.929104   5372.0  3.332906e-07      1.0  448.0  0.000004\n",
      "CPU times: user 2min 21s, sys: 22.4 s, total: 2min 43s\n",
      "Wall time: 2min 54s\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>train_loss</th>\n",
       "      <th>train_acc</th>\n",
       "      <th>n_train</th>\n",
       "      <th>val_loss</th>\n",
       "      <th>val_acc</th>\n",
       "      <th>n_val</th>\n",
       "      <th>lr</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>epoch</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.199804</td>\n",
       "      <td>0.102612</td>\n",
       "      <td>5372</td>\n",
       "      <td>1.145504e-01</td>\n",
       "      <td>0.472727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000060</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.117295</td>\n",
       "      <td>0.388246</td>\n",
       "      <td>5372</td>\n",
       "      <td>3.938448e-02</td>\n",
       "      <td>0.843182</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000060</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.066997</td>\n",
       "      <td>0.625373</td>\n",
       "      <td>5372</td>\n",
       "      <td>1.773462e-02</td>\n",
       "      <td>0.934091</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000060</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.051026</td>\n",
       "      <td>0.708582</td>\n",
       "      <td>5372</td>\n",
       "      <td>1.262682e-02</td>\n",
       "      <td>0.945455</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000060</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.042064</td>\n",
       "      <td>0.759701</td>\n",
       "      <td>5372</td>\n",
       "      <td>7.379494e-03</td>\n",
       "      <td>0.975000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000060</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.034712</td>\n",
       "      <td>0.793470</td>\n",
       "      <td>5372</td>\n",
       "      <td>7.104102e-03</td>\n",
       "      <td>0.965909</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000060</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.030932</td>\n",
       "      <td>0.815485</td>\n",
       "      <td>5372</td>\n",
       "      <td>2.055051e-03</td>\n",
       "      <td>0.979545</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000060</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0.025177</td>\n",
       "      <td>0.846642</td>\n",
       "      <td>5372</td>\n",
       "      <td>1.502132e-03</td>\n",
       "      <td>0.993182</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000060</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.025183</td>\n",
       "      <td>0.850187</td>\n",
       "      <td>5372</td>\n",
       "      <td>1.170390e-03</td>\n",
       "      <td>0.995455</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000060</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.025113</td>\n",
       "      <td>0.848694</td>\n",
       "      <td>5372</td>\n",
       "      <td>1.879398e-03</td>\n",
       "      <td>0.995455</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000060</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0.023708</td>\n",
       "      <td>0.860075</td>\n",
       "      <td>5372</td>\n",
       "      <td>2.283715e-03</td>\n",
       "      <td>0.984091</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000060</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>0.018856</td>\n",
       "      <td>0.876866</td>\n",
       "      <td>5372</td>\n",
       "      <td>1.606588e-05</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>0.016996</td>\n",
       "      <td>0.888993</td>\n",
       "      <td>5372</td>\n",
       "      <td>1.991304e-05</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>0.015972</td>\n",
       "      <td>0.898134</td>\n",
       "      <td>5372</td>\n",
       "      <td>5.376311e-06</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0.016419</td>\n",
       "      <td>0.896269</td>\n",
       "      <td>5372</td>\n",
       "      <td>7.298137e-06</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>0.014811</td>\n",
       "      <td>0.909701</td>\n",
       "      <td>5372</td>\n",
       "      <td>4.395133e-05</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>0.016481</td>\n",
       "      <td>0.897015</td>\n",
       "      <td>5372</td>\n",
       "      <td>6.370148e-04</td>\n",
       "      <td>0.997727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>0.015402</td>\n",
       "      <td>0.901493</td>\n",
       "      <td>5372</td>\n",
       "      <td>1.025263e-04</td>\n",
       "      <td>0.997727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>0.013993</td>\n",
       "      <td>0.907276</td>\n",
       "      <td>5372</td>\n",
       "      <td>1.344587e-05</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>0.013496</td>\n",
       "      <td>0.911194</td>\n",
       "      <td>5372</td>\n",
       "      <td>6.435510e-06</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>0.013600</td>\n",
       "      <td>0.911381</td>\n",
       "      <td>5372</td>\n",
       "      <td>7.953949e-04</td>\n",
       "      <td>0.997727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>0.013086</td>\n",
       "      <td>0.914179</td>\n",
       "      <td>5372</td>\n",
       "      <td>9.326528e-06</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>0.012593</td>\n",
       "      <td>0.915299</td>\n",
       "      <td>5372</td>\n",
       "      <td>5.242742e-05</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>0.014211</td>\n",
       "      <td>0.910261</td>\n",
       "      <td>5372</td>\n",
       "      <td>2.584969e-04</td>\n",
       "      <td>0.997727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>0.011789</td>\n",
       "      <td>0.919403</td>\n",
       "      <td>5372</td>\n",
       "      <td>2.376859e-04</td>\n",
       "      <td>0.997727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>0.011660</td>\n",
       "      <td>0.921269</td>\n",
       "      <td>5372</td>\n",
       "      <td>1.307089e-04</td>\n",
       "      <td>0.997727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>0.012474</td>\n",
       "      <td>0.914179</td>\n",
       "      <td>5372</td>\n",
       "      <td>1.289003e-04</td>\n",
       "      <td>0.997727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>0.012148</td>\n",
       "      <td>0.920522</td>\n",
       "      <td>5372</td>\n",
       "      <td>1.509231e-04</td>\n",
       "      <td>0.997727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>0.010933</td>\n",
       "      <td>0.926679</td>\n",
       "      <td>5372</td>\n",
       "      <td>8.473287e-07</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>0.011522</td>\n",
       "      <td>0.923507</td>\n",
       "      <td>5372</td>\n",
       "      <td>1.156212e-05</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>0.011791</td>\n",
       "      <td>0.925187</td>\n",
       "      <td>5372</td>\n",
       "      <td>2.362675e-04</td>\n",
       "      <td>0.997727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>0.011458</td>\n",
       "      <td>0.922015</td>\n",
       "      <td>5372</td>\n",
       "      <td>5.158910e-07</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>0.011378</td>\n",
       "      <td>0.922015</td>\n",
       "      <td>5372</td>\n",
       "      <td>8.632589e-05</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>0.010526</td>\n",
       "      <td>0.925746</td>\n",
       "      <td>5372</td>\n",
       "      <td>2.282475e-07</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000004</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>0.010228</td>\n",
       "      <td>0.931343</td>\n",
       "      <td>5372</td>\n",
       "      <td>4.299960e-07</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000004</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>0.011143</td>\n",
       "      <td>0.922201</td>\n",
       "      <td>5372</td>\n",
       "      <td>2.418208e-07</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000004</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>0.010649</td>\n",
       "      <td>0.926866</td>\n",
       "      <td>5372</td>\n",
       "      <td>1.234686e-06</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000004</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>0.010454</td>\n",
       "      <td>0.928918</td>\n",
       "      <td>5372</td>\n",
       "      <td>1.563356e-07</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000004</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>0.010814</td>\n",
       "      <td>0.925746</td>\n",
       "      <td>5372</td>\n",
       "      <td>2.169874e-07</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000004</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>0.010376</td>\n",
       "      <td>0.929104</td>\n",
       "      <td>5372</td>\n",
       "      <td>3.332906e-07</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000004</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       train_loss  train_acc  n_train      val_loss   val_acc  n_val        lr\n",
       "epoch                                                                         \n",
       "1        0.199804   0.102612     5372  1.145504e-01  0.472727    448  0.000060\n",
       "2        0.117295   0.388246     5372  3.938448e-02  0.843182    448  0.000060\n",
       "3        0.066997   0.625373     5372  1.773462e-02  0.934091    448  0.000060\n",
       "4        0.051026   0.708582     5372  1.262682e-02  0.945455    448  0.000060\n",
       "5        0.042064   0.759701     5372  7.379494e-03  0.975000    448  0.000060\n",
       "6        0.034712   0.793470     5372  7.104102e-03  0.965909    448  0.000060\n",
       "7        0.030932   0.815485     5372  2.055051e-03  0.979545    448  0.000060\n",
       "8        0.025177   0.846642     5372  1.502132e-03  0.993182    448  0.000060\n",
       "9        0.025183   0.850187     5372  1.170390e-03  0.995455    448  0.000060\n",
       "10       0.025113   0.848694     5372  1.879398e-03  0.995455    448  0.000060\n",
       "11       0.023708   0.860075     5372  2.283715e-03  0.984091    448  0.000060\n",
       "12       0.018856   0.876866     5372  1.606588e-05  1.000000    448  0.000024\n",
       "13       0.016996   0.888993     5372  1.991304e-05  1.000000    448  0.000024\n",
       "14       0.015972   0.898134     5372  5.376311e-06  1.000000    448  0.000024\n",
       "15       0.016419   0.896269     5372  7.298137e-06  1.000000    448  0.000024\n",
       "16       0.014811   0.909701     5372  4.395133e-05  1.000000    448  0.000024\n",
       "17       0.016481   0.897015     5372  6.370148e-04  0.997727    448  0.000024\n",
       "18       0.015402   0.901493     5372  1.025263e-04  0.997727    448  0.000024\n",
       "19       0.013993   0.907276     5372  1.344587e-05  1.000000    448  0.000024\n",
       "20       0.013496   0.911194     5372  6.435510e-06  1.000000    448  0.000024\n",
       "21       0.013600   0.911381     5372  7.953949e-04  0.997727    448  0.000024\n",
       "22       0.013086   0.914179     5372  9.326528e-06  1.000000    448  0.000024\n",
       "23       0.012593   0.915299     5372  5.242742e-05  1.000000    448  0.000010\n",
       "24       0.014211   0.910261     5372  2.584969e-04  0.997727    448  0.000010\n",
       "25       0.011789   0.919403     5372  2.376859e-04  0.997727    448  0.000010\n",
       "26       0.011660   0.921269     5372  1.307089e-04  0.997727    448  0.000010\n",
       "27       0.012474   0.914179     5372  1.289003e-04  0.997727    448  0.000010\n",
       "28       0.012148   0.920522     5372  1.509231e-04  0.997727    448  0.000010\n",
       "29       0.010933   0.926679     5372  8.473287e-07  1.000000    448  0.000010\n",
       "30       0.011522   0.923507     5372  1.156212e-05  1.000000    448  0.000010\n",
       "31       0.011791   0.925187     5372  2.362675e-04  0.997727    448  0.000010\n",
       "32       0.011458   0.922015     5372  5.158910e-07  1.000000    448  0.000010\n",
       "33       0.011378   0.922015     5372  8.632589e-05  1.000000    448  0.000010\n",
       "34       0.010526   0.925746     5372  2.282475e-07  1.000000    448  0.000004\n",
       "35       0.010228   0.931343     5372  4.299960e-07  1.000000    448  0.000004\n",
       "36       0.011143   0.922201     5372  2.418208e-07  1.000000    448  0.000004\n",
       "37       0.010649   0.926866     5372  1.234686e-06  1.000000    448  0.000004\n",
       "38       0.010454   0.928918     5372  1.563356e-07  1.000000    448  0.000004\n",
       "39       0.010814   0.925746     5372  2.169874e-07  1.000000    448  0.000004\n",
       "40       0.010376   0.929104     5372  3.332906e-07  1.000000    448  0.000004"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "trainer.train(40)\n",
    "trainer.complete_results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "e27d6d98-66a8-4a6b-a378-588a34a18822",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       test_loss  test_acc  n_test        lr\n",
      "epoch                                       \n",
      "40      0.000146  0.998864   896.0  0.000004\n",
      "CPU times: user 246 ms, sys: 258 ms, total: 504 ms\n",
      "Wall time: 627 ms\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>test_loss</th>\n",
       "      <th>test_acc</th>\n",
       "      <th>n_test</th>\n",
       "      <th>lr</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>epoch</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>0.000146</td>\n",
       "      <td>0.998864</td>\n",
       "      <td>896.0</td>\n",
       "      <td>0.000004</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       test_loss  test_acc  n_test        lr\n",
       "epoch                                       \n",
       "40      0.000146  0.998864   896.0  0.000004"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "trainer.test()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ed35c728-df73-44b5-b315-5fde6dd1cfbf",
   "metadata": {},
   "source": [
    "Despite the vary number of classes, the model proves very accurate with a near 100% balanced accuracy in character predictions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "17d6e727-e044-413e-aedd-a87b6d2bdcd8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1400x700 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, [ax1, ax2] = common_plot_setup(style=\"fivethirtyeight\", font=\"DejaVu Sans\", figsize=(14, 7), rows=1, cols=2)\n",
    "trainer.complete_results[[\"train_loss\", \"val_loss\"]].plot(ax=ax1, logy=True)\n",
    "trainer.complete_results[[\"train_acc\", \"val_acc\"]].plot(ax=ax2, logy=False)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "22ca27f5-8718-4f80-86cf-e8800360a547",
   "metadata": {},
   "source": [
    "During this round of training the model only misclassified a single character. By examining the scaled outputs of the classification layer it is apparent that the model was mostly right but messed up on the character's diacritic mistaking the circular handakuten for the small pair of strokes that make up a dakuten. Even though stroke count was distinctly encoded in the rgb values it wasn't enough to be identified by the model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "f9b34cd8-ab88-4b8f-b582-7c62c9d743b8",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Characters missed: 1\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>label</th>\n",
       "      <th>probability</th>\n",
       "      <th>char_id</th>\n",
       "      <th>character</th>\n",
       "      <th>romanization</th>\n",
       "      <th>stroke_count</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <th>label</th>\n",
       "      <th>char_id</th>\n",
       "      <th>character</th>\n",
       "      <th>romanization</th>\n",
       "      <th>indices</th>\n",
       "      <th>stroke_count</th>\n",
       "      <th>strokes_found</th>\n",
       "      <th>pred</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">1</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">hir_bya</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">bya</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">びゃ</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">bya</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">0</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">6</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">6.0</th>\n",
       "      <th>1</th>\n",
       "      <td>hir_pya</td>\n",
       "      <td>0.925672</td>\n",
       "      <td>pya</td>\n",
       "      <td>ぴゃ</td>\n",
       "      <td>pya</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>hir_bya</td>\n",
       "      <td>0.074299</td>\n",
       "      <td>bya</td>\n",
       "      <td>びゃ</td>\n",
       "      <td>bya</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hir_kya</td>\n",
       "      <td>0.000022</td>\n",
       "      <td>kya</td>\n",
       "      <td>きゃ</td>\n",
       "      <td>kya</td>\n",
       "      <td>6 or 7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>hir_zya</td>\n",
       "      <td>0.000004</td>\n",
       "      <td>zya</td>\n",
       "      <td>じゃ</td>\n",
       "      <td>ja</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hir_e</td>\n",
       "      <td>0.000002</td>\n",
       "      <td>e</td>\n",
       "      <td>え</td>\n",
       "      <td>e</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                                        label  \\\n",
       "count label   char_id character romanization indices stroke_count strokes_found pred            \n",
       "1     hir_bya bya     びゃ        bya          0       6            6.0           1     hir_pya   \n",
       "                                                                                2     hir_bya   \n",
       "                                                                                3     hir_kya   \n",
       "                                                                                4     hir_zya   \n",
       "                                                                                5       hir_e   \n",
       "\n",
       "                                                                                      probability  \\\n",
       "count label   char_id character romanization indices stroke_count strokes_found pred                \n",
       "1     hir_bya bya     びゃ        bya          0       6            6.0           1        0.925672   \n",
       "                                                                                2        0.074299   \n",
       "                                                                                3        0.000022   \n",
       "                                                                                4        0.000004   \n",
       "                                                                                5        0.000002   \n",
       "\n",
       "                                                                                     char_id  \\\n",
       "count label   char_id character romanization indices stroke_count strokes_found pred           \n",
       "1     hir_bya bya     びゃ        bya          0       6            6.0           1        pya   \n",
       "                                                                                2        bya   \n",
       "                                                                                3        kya   \n",
       "                                                                                4        zya   \n",
       "                                                                                5          e   \n",
       "\n",
       "                                                                                     character  \\\n",
       "count label   char_id character romanization indices stroke_count strokes_found pred             \n",
       "1     hir_bya bya     びゃ        bya          0       6            6.0           1           ぴゃ   \n",
       "                                                                                2           びゃ   \n",
       "                                                                                3           きゃ   \n",
       "                                                                                4           じゃ   \n",
       "                                                                                5            え   \n",
       "\n",
       "                                                                                     romanization  \\\n",
       "count label   char_id character romanization indices stroke_count strokes_found pred                \n",
       "1     hir_bya bya     びゃ        bya          0       6            6.0           1             pya   \n",
       "                                                                                2             bya   \n",
       "                                                                                3             kya   \n",
       "                                                                                4              ja   \n",
       "                                                                                5               e   \n",
       "\n",
       "                                                                                     stroke_count  \n",
       "count label   char_id character romanization indices stroke_count strokes_found pred               \n",
       "1     hir_bya bya     びゃ        bya          0       6            6.0           1               5  \n",
       "                                                                                2               6  \n",
       "                                                                                3          6 or 7  \n",
       "                                                                                4               6  \n",
       "                                                                                5               2  "
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "misses = IncorrectCharacters(trainer, transformed_dataset, data_paths.get_classes())\n",
    "print(\"Characters missed:\", len(misses))\n",
    "misses.overview(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0a02df26-6a26-41d4-a9e2-898427db4941",
   "metadata": {},
   "source": [
    "## Model 2: Simple LSTM"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dad57326-1371-45bb-9234-8630c1138329",
   "metadata": {},
   "source": [
    "The next model tested is a very basic LSTM. What is interesting about this dataset is that it contains two sequential dimensions: the points drawn to form a stroke and the strokes drawn to form a character. A basic RNN is traditionally meant for one of these (but this notebook also contains a few architectures that feed the temporal points from one into another). The model used in the next few cells expands the feature dimension to accomodate all the parameters for each stroke side by side (batch, stroke, points, features) -> (batch, points, features for each stroke in parallel)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "332b07ed-0de9-41bb-ba40-62affe2e7313",
   "metadata": {},
   "outputs": [],
   "source": [
    "class StrokeMelter:\n",
    "\n",
    "    def __call__(self, sample):\n",
    "        return sample.reshape((-1, sample.shape[-1]))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e5052f35-325d-4856-b904-0f7a7e728b8b",
   "metadata": {},
   "source": [
    "Before defining the model, some aggregate values will be taken to help normalize the outputs to a range between -1 and 1 followed by a shift to the new mean (for the features that are significantly off center)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "c1c2ab79-c2ea-4532-b6dd-55b4714ee4e8",
   "metadata": {},
   "outputs": [],
   "source": [
    "aggregator = StrokeDatasetAggregator(data_paths, transforms.Compose([ToBSplines(120), # Replacing the irregular sequences of points with the uniform counts fitted to their b-splines\n",
    "                                                                     ArrayToTensor(),\n",
    "                                                                     ExtractAngles(include_magnitudes=False), # Extracts the sine and cosine values between the x and y coordinate pairs as additional features\n",
    "                                                                     StrokeMelter() # Dissolves the stroke dimension so that the aggregates across all points and strokes.\n",
    "                                                                    ]), None\n",
    "                                    )"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "84e06667-5c06-493b-8900-cbcbda69c72d",
   "metadata": {},
   "source": [
    "As expected the actual range of the features show very little deviation from -1 and 1. In the case of the coordinates, those are the bounds of the normalized canvas. It seems like at least one of the points is outside of this range (presumably a small shift from the b-spline fitting). The angles between points already are true to the maximimums and minimums of the sine and cosine functions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "940ace0b-96af-4b49-a048-26775e4d11a7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min: tensor([[-0.9900, -0.9857, -1.0000, -1.0000]])\n",
      "Max: tensor([[1.0225, 0.9961, 1.0000, 1.0000]])\n"
     ]
    }
   ],
   "source": [
    "old_min, old_max = aggregator.get_min_max(dim=0, keepdim=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "63998405-8ae2-4532-9e12-6169b308e082",
   "metadata": {},
   "source": [
    "In order to identify any skews in the inputs, the mean across all features is calculated. The angular data it appears is heavily skewed due to the conventions of proper stroke order."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "e4807537-b91c-430a-a8ee-963ba64f5085",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean: tensor([[ 0.0877,  0.1766,  0.4025, -0.4613]])\n",
      "STD: tensor([[0.4305, 0.3686, 0.5654, 0.5527]])\n"
     ]
    }
   ],
   "source": [
    "means, _ = aggregator.get_mean_std(dim=0, keepdim=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fe9ff9e3-737f-45b2-9762-9c43ec6997f5",
   "metadata": {},
   "source": [
    "There is also a smaller skew present in the coordinates. Because the dataset is far from comprehensive of the Japanese writing system and is lacking in diversity of handwriting styles, I've chosen to leave the positioning as is while only recentering the angular values."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "08d78f3a-aa0b-4489-8a16-b2ec1433d1b5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([[ 0.0710,  0.1730,  0.4025, -0.4613]])\n",
      "tensor([[[ 0.0000,  0.0000,  0.4025, -0.4613]]])\n"
     ]
    }
   ],
   "source": [
    "std_means = (means - old_min) / (old_max - old_min) * 2 - 1\n",
    "print(std_means)\n",
    "std_shift = f.pad(std_means[:,2:4], (2, 0)).unsqueeze(0)\n",
    "old_min, old_max = old_min.unsqueeze(0), old_max.unsqueeze(0)\n",
    "print(std_shift)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5a229348-c090-41de-bcd8-5e4b1ca93fb7",
   "metadata": {},
   "source": [
    "The follwoing cell defines the LSTM as well as the transformation required to make the inputs compatible with this architecture."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "a440c458-cc0f-45da-ac5d-2376a7899974",
   "metadata": {},
   "outputs": [],
   "source": [
    "class StrokeSquasher:\n",
    "    def __call__(self, sample):\n",
    "        n_strokes, _, n_features = sample.shape\n",
    "        sample = sample.permute(1, 0, 2)\n",
    "        return sample.reshape(-1, n_strokes * n_features)\n",
    "\n",
    "class CharacterLSTM(nn.Module):\n",
    "    def __init__(self, lstm_features, lstm_hidden, dense_hidden, num_classes):\n",
    "        super().__init__()\n",
    "        self._lstm_hidden = lstm_hidden\n",
    "        self.lstm = nn.LSTM(lstm_features, lstm_hidden, num_layers=1, batch_first=True)\n",
    "        self.linear1 = nn.Linear(lstm_hidden, dense_hidden)\n",
    "        self.linear2 = nn.Linear(dense_hidden, num_classes)\n",
    "        self.dropout = nn.Dropout(p = 0.1)\n",
    "        self.relu = nn.ReLU()\n",
    "        self.lstm_h = (torch.zeros((0)), torch.zeros((0)))\n",
    "        self.dummy_param = nn.Parameter(torch.empty(0))\n",
    "\n",
    "    def _init_hidden_state(self, previous, x, hidden_size):\n",
    "        if tuple(previous.shape) != (1, x.shape[0], hidden_size):\n",
    "            device = self.dummy_param.device\n",
    "            return torch.zeros(1, x.shape[0], hidden_size).to(device)\n",
    "        return previous\n",
    "\n",
    "    def _init_lstm_hidden_states(self, x):\n",
    "        h = self._init_hidden_state(self.lstm_h[0], x, self._lstm_hidden)\n",
    "        if self.lstm_h[0] is not h:\n",
    "            self.lstm_h = (h, h.clone())\n",
    "\n",
    "    def _lstm_forward(self, x):\n",
    "        self._init_lstm_hidden_states(x)\n",
    "        out, (h, c) = self.lstm(x, self.lstm_h)\n",
    "        out = out[:,-1].view(-1, self._lstm_hidden)\n",
    "        return out, (h, c)\n",
    "\n",
    "    def forward(self, inputs):\n",
    "        out, (h, c) = self._lstm_forward(inputs)\n",
    "        out = self.dropout(out)\n",
    "        out = self.linear1(out)\n",
    "        out = self.relu(out)\n",
    "        out = self.linear2(out)\n",
    "        return out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "7abc15a9-08de-43d3-b362-525cb32e4926",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'train': <torch.utils.data.dataloader.DataLoader at 0x7ffaef7e46d0>,\n",
       " 'val': <torch.utils.data.dataloader.DataLoader at 0x7ffa1bf79ad0>,\n",
       " 'test': <torch.utils.data.dataloader.DataLoader at 0x7ffa16189890>}"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_transforms = [ToBSplines(120), ArrayToTensor(),\n",
    "                   ExtractAngles(include_magnitudes=False),\n",
    "                   InputMinMaxTransformer(None, old_min, old_max, new_min = -1), # Normalizes against the aggregated dataset to a minmax range of -1 to 1.\n",
    "                   InputRecenter(None, std_shift), # Recenters the angular data against the aggregated mean.\n",
    "                  ]\n",
    "end_transforms = [EmptyStrokePadder(stroke_count, 0, 0), # Adds padding to give the stroke dimension a consistent size of 10.\n",
    "                  StrokeSquasher() # Moves stroke information to the feature dimension for the LSTM.\n",
    "                 ]\n",
    "# Training dataset will be augmented by random transformations.\n",
    "train_transforms = transforms.Compose(test_transforms + [InputGaussianNoise(None, .1)] + end_transforms)\n",
    "test_transforms = transforms.Compose(test_transforms + end_transforms)\n",
    "transformed_dataset = StrokeDataset(data_paths.get_data(), transform=test_transforms)\n",
    "split_keys = {\"train\": train_transforms, \"val\": test_transforms, \"test\": test_transforms}\n",
    "split_datapaths = dict(zip(split_keys, data_paths.train_val_test_split(train_size=.7, test_size=.2, random_state=SEED)))\n",
    "datasets = {key: StrokeDataset(split_datapaths[key], transform=value) for key, value in split_keys.items()}\n",
    "dataloaders = {key: torch.utils.data.DataLoader(datasets[key], batch_size=batch_size, shuffle=True, num_workers=5, drop_last=True) for key in split_keys.keys()}\n",
    "dataloaders"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "deba364e-974b-4673-9700-fb5e757fdf62",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "108\n"
     ]
    }
   ],
   "source": [
    "# All features must be repeated for each stroke.\n",
    "lstm_features = stroke_count * 4\n",
    "lstm_hidden_size = 400\n",
    "dense_hidden_size = 46\n",
    "learning_rate=0.003\n",
    "n_classes = np.unique(full_dataset.labels).size\n",
    "print(n_classes)\n",
    "\n",
    "model = CharacterLSTM(lstm_features, lstm_hidden_size, dense_hidden_size, n_classes)\n",
    "criterion = nn.CrossEntropyLoss()\n",
    "optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)\n",
    "scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=11, gamma=0.4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "d3281b9e-e29c-4af6-bd93-c95dd00c3077",
   "metadata": {},
   "outputs": [],
   "source": [
    "trainer = CharacterTrainer(model, optimizer, criterion, dataloaders, scheduler=scheduler, balance_acc=True, single_input=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "3a6db760-45c5-407a-bcc5-5ff20c48282f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "1        0.131719   0.317018   5372.0   0.05772  0.663814  448.0  0.003\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/mambaforge/envs/choubenkyo/lib/python3.11/site-packages/sklearn/metrics/_classification.py:2399: UserWarning: y_pred contains classes not in y_true\n",
      "  warnings.warn(\"y_pred contains classes not in y_true\")\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "2        0.034454   0.783364   5372.0  0.022212  0.876033  448.0  0.003\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/mambaforge/envs/choubenkyo/lib/python3.11/site-packages/sklearn/metrics/_classification.py:2399: UserWarning: y_pred contains classes not in y_true\n",
      "  warnings.warn(\"y_pred contains classes not in y_true\")\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "3        0.020383   0.867454   5372.0  0.011855  0.931825  448.0  0.003\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "4        0.011051   0.924723   5372.0  0.011779  0.933333  448.0  0.003\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/mambaforge/envs/choubenkyo/lib/python3.11/site-packages/sklearn/metrics/_classification.py:2399: UserWarning: y_pred contains classes not in y_true\n",
      "  warnings.warn(\"y_pred contains classes not in y_true\")\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "5        0.008658    0.93837   5372.0  0.006413  0.962343  448.0  0.003\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "6        0.006137   0.958755   5372.0  0.005516  0.974113  448.0  0.003\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "7        0.005034   0.963776   5372.0  0.007417  0.955256  448.0  0.003\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/mambaforge/envs/choubenkyo/lib/python3.11/site-packages/sklearn/metrics/_classification.py:2399: UserWarning: y_pred contains classes not in y_true\n",
      "  warnings.warn(\"y_pred contains classes not in y_true\")\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "8        0.005758   0.959959   5372.0  0.006607  0.963533  448.0  0.003\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "9        0.006264     0.9601   5372.0  0.005119  0.967958  448.0  0.003\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "10       0.004854   0.965617   5372.0  0.015833  0.961557  448.0  0.003\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "11       0.005649   0.962568   5372.0  0.004398  0.986792  448.0  0.003\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/mambaforge/envs/choubenkyo/lib/python3.11/site-packages/sklearn/metrics/_classification.py:2399: UserWarning: y_pred contains classes not in y_true\n",
      "  warnings.warn(\"y_pred contains classes not in y_true\")\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "12       0.001832   0.989184   5372.0  0.003251  0.988481  448.0  0.0012\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "13       0.000988   0.994979   5372.0  0.002767  0.987063  448.0  0.0012\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "14        0.00082   0.995996   5372.0  0.002817  0.989297  448.0  0.0012\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/mambaforge/envs/choubenkyo/lib/python3.11/site-packages/sklearn/metrics/_classification.py:2399: UserWarning: y_pred contains classes not in y_true\n",
      "  warnings.warn(\"y_pred contains classes not in y_true\")\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "15       0.000826   0.995353   5372.0  0.002424  0.991588  448.0  0.0012\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "16       0.000515   0.996943   5372.0  0.001929  0.996934  448.0  0.0012\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "17       0.000412   0.998025   5372.0  0.002534  0.993789  448.0  0.0012\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "18       0.001193   0.991712   5372.0  0.002982  0.988848  448.0  0.0012\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "19       0.000887   0.992551   5372.0  0.002051  0.992217  448.0  0.0012\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "20       0.000793   0.993223   5372.0  0.003178  0.989522  448.0  0.0012\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/mambaforge/envs/choubenkyo/lib/python3.11/site-packages/sklearn/metrics/_classification.py:2399: UserWarning: y_pred contains classes not in y_true\n",
      "  warnings.warn(\"y_pred contains classes not in y_true\")\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "21       0.000874   0.992792   5372.0  0.001943  0.994239  448.0  0.0012\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "22       0.004892    0.96881   5372.0  0.003582  0.987186  448.0  0.0012\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "23       0.001487   0.988858   5372.0  0.002217  0.988982  448.0  0.00048\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "24       0.000915   0.994382   5372.0  0.002288  0.990555  448.0  0.00048\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "25        0.00079   0.994315   5372.0  0.002182  0.988567  448.0  0.00048\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "26       0.000577   0.996377   5372.0  0.001856  0.990083  448.0  0.00048\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "27       0.000577   0.996901   5372.0  0.002426  0.988589  448.0  0.00048\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "28       0.000599   0.994965   5372.0  0.001855  0.991903  448.0  0.00048\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "29       0.000496   0.996243   5372.0  0.002147  0.987186  448.0  0.00048\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "30       0.000677    0.99559   5372.0  0.001686  0.992352  448.0  0.00048\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "31       0.000545    0.99545   5372.0  0.001842  0.993639  448.0  0.00048\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "32       0.000563   0.996313   5372.0     0.002  0.995047  448.0  0.00048\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "33       0.000526   0.996408   5372.0  0.002131  0.993699  448.0  0.00048\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val        lr\n",
      "epoch                                                                     \n",
      "34       0.000247   0.998544   5372.0  0.002118  0.993228  448.0  0.000192\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val        lr\n",
      "epoch                                                                     \n",
      "35       0.000384   0.997725   5372.0  0.001959  0.995047  448.0  0.000192\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val        lr\n",
      "epoch                                                                     \n",
      "36       0.000143   0.999669   5372.0  0.001535  0.994879  448.0  0.000192\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val        lr\n",
      "epoch                                                                     \n",
      "37       0.000423   0.996612   5372.0  0.001824  0.993228  448.0  0.000192\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val        lr\n",
      "epoch                                                                     \n",
      "38       0.000379   0.996707   5372.0  0.001912  0.993531  448.0  0.000192\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val        lr\n",
      "epoch                                                                     \n",
      "39       0.000332   0.996936   5372.0  0.001881  0.993531  448.0  0.000192\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val        lr\n",
      "epoch                                                                     \n",
      "40       0.000305   0.997826   5372.0  0.001928  0.995047  448.0  0.000192\n",
      "CPU times: user 1min 35s, sys: 23.6 s, total: 1min 58s\n",
      "Wall time: 2min 11s\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>train_loss</th>\n",
       "      <th>train_acc</th>\n",
       "      <th>n_train</th>\n",
       "      <th>val_loss</th>\n",
       "      <th>val_acc</th>\n",
       "      <th>n_val</th>\n",
       "      <th>lr</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>epoch</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.131719</td>\n",
       "      <td>0.317018</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.057720</td>\n",
       "      <td>0.663814</td>\n",
       "      <td>448</td>\n",
       "      <td>0.003000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.034454</td>\n",
       "      <td>0.783364</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.022212</td>\n",
       "      <td>0.876033</td>\n",
       "      <td>448</td>\n",
       "      <td>0.003000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.020383</td>\n",
       "      <td>0.867454</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.011855</td>\n",
       "      <td>0.931825</td>\n",
       "      <td>448</td>\n",
       "      <td>0.003000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.011051</td>\n",
       "      <td>0.924723</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.011779</td>\n",
       "      <td>0.933333</td>\n",
       "      <td>448</td>\n",
       "      <td>0.003000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.008658</td>\n",
       "      <td>0.938370</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.006413</td>\n",
       "      <td>0.962343</td>\n",
       "      <td>448</td>\n",
       "      <td>0.003000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.006137</td>\n",
       "      <td>0.958755</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.005516</td>\n",
       "      <td>0.974113</td>\n",
       "      <td>448</td>\n",
       "      <td>0.003000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.005034</td>\n",
       "      <td>0.963776</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.007417</td>\n",
       "      <td>0.955256</td>\n",
       "      <td>448</td>\n",
       "      <td>0.003000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0.005758</td>\n",
       "      <td>0.959959</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.006607</td>\n",
       "      <td>0.963533</td>\n",
       "      <td>448</td>\n",
       "      <td>0.003000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.006264</td>\n",
       "      <td>0.960100</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.005119</td>\n",
       "      <td>0.967958</td>\n",
       "      <td>448</td>\n",
       "      <td>0.003000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.004854</td>\n",
       "      <td>0.965617</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.015833</td>\n",
       "      <td>0.961557</td>\n",
       "      <td>448</td>\n",
       "      <td>0.003000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0.005649</td>\n",
       "      <td>0.962568</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.004398</td>\n",
       "      <td>0.986792</td>\n",
       "      <td>448</td>\n",
       "      <td>0.003000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>0.001832</td>\n",
       "      <td>0.989184</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.003251</td>\n",
       "      <td>0.988481</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>0.000988</td>\n",
       "      <td>0.994979</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.002767</td>\n",
       "      <td>0.987063</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>0.000820</td>\n",
       "      <td>0.995996</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.002817</td>\n",
       "      <td>0.989297</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0.000826</td>\n",
       "      <td>0.995353</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.002424</td>\n",
       "      <td>0.991588</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>0.000515</td>\n",
       "      <td>0.996943</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.001929</td>\n",
       "      <td>0.996934</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>0.000412</td>\n",
       "      <td>0.998025</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.002534</td>\n",
       "      <td>0.993789</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>0.001193</td>\n",
       "      <td>0.991712</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.002982</td>\n",
       "      <td>0.988848</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>0.000887</td>\n",
       "      <td>0.992551</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.002051</td>\n",
       "      <td>0.992217</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>0.000793</td>\n",
       "      <td>0.993223</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.003178</td>\n",
       "      <td>0.989522</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>0.000874</td>\n",
       "      <td>0.992792</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.001943</td>\n",
       "      <td>0.994239</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>0.004892</td>\n",
       "      <td>0.968810</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.003582</td>\n",
       "      <td>0.987186</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>0.001487</td>\n",
       "      <td>0.988858</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.002217</td>\n",
       "      <td>0.988982</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000480</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>0.000915</td>\n",
       "      <td>0.994382</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.002288</td>\n",
       "      <td>0.990555</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000480</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>0.000790</td>\n",
       "      <td>0.994315</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.002182</td>\n",
       "      <td>0.988567</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000480</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>0.000577</td>\n",
       "      <td>0.996377</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.001856</td>\n",
       "      <td>0.990083</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000480</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>0.000577</td>\n",
       "      <td>0.996901</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.002426</td>\n",
       "      <td>0.988589</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000480</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>0.000599</td>\n",
       "      <td>0.994965</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.001855</td>\n",
       "      <td>0.991903</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000480</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>0.000496</td>\n",
       "      <td>0.996243</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.002147</td>\n",
       "      <td>0.987186</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000480</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>0.000677</td>\n",
       "      <td>0.995590</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.001686</td>\n",
       "      <td>0.992352</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000480</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>0.000545</td>\n",
       "      <td>0.995450</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.001842</td>\n",
       "      <td>0.993639</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000480</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>0.000563</td>\n",
       "      <td>0.996313</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.002000</td>\n",
       "      <td>0.995047</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000480</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>0.000526</td>\n",
       "      <td>0.996408</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.002131</td>\n",
       "      <td>0.993699</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000480</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>0.000247</td>\n",
       "      <td>0.998544</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.002118</td>\n",
       "      <td>0.993228</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000192</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>0.000384</td>\n",
       "      <td>0.997725</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.001959</td>\n",
       "      <td>0.995047</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000192</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>0.000143</td>\n",
       "      <td>0.999669</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.001535</td>\n",
       "      <td>0.994879</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000192</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>0.000423</td>\n",
       "      <td>0.996612</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.001824</td>\n",
       "      <td>0.993228</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000192</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>0.000379</td>\n",
       "      <td>0.996707</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.001912</td>\n",
       "      <td>0.993531</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000192</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>0.000332</td>\n",
       "      <td>0.996936</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.001881</td>\n",
       "      <td>0.993531</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000192</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>0.000305</td>\n",
       "      <td>0.997826</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.001928</td>\n",
       "      <td>0.995047</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000192</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       train_loss  train_acc  n_train  val_loss   val_acc  n_val        lr\n",
       "epoch                                                                     \n",
       "1        0.131719   0.317018     5372  0.057720  0.663814    448  0.003000\n",
       "2        0.034454   0.783364     5372  0.022212  0.876033    448  0.003000\n",
       "3        0.020383   0.867454     5372  0.011855  0.931825    448  0.003000\n",
       "4        0.011051   0.924723     5372  0.011779  0.933333    448  0.003000\n",
       "5        0.008658   0.938370     5372  0.006413  0.962343    448  0.003000\n",
       "6        0.006137   0.958755     5372  0.005516  0.974113    448  0.003000\n",
       "7        0.005034   0.963776     5372  0.007417  0.955256    448  0.003000\n",
       "8        0.005758   0.959959     5372  0.006607  0.963533    448  0.003000\n",
       "9        0.006264   0.960100     5372  0.005119  0.967958    448  0.003000\n",
       "10       0.004854   0.965617     5372  0.015833  0.961557    448  0.003000\n",
       "11       0.005649   0.962568     5372  0.004398  0.986792    448  0.003000\n",
       "12       0.001832   0.989184     5372  0.003251  0.988481    448  0.001200\n",
       "13       0.000988   0.994979     5372  0.002767  0.987063    448  0.001200\n",
       "14       0.000820   0.995996     5372  0.002817  0.989297    448  0.001200\n",
       "15       0.000826   0.995353     5372  0.002424  0.991588    448  0.001200\n",
       "16       0.000515   0.996943     5372  0.001929  0.996934    448  0.001200\n",
       "17       0.000412   0.998025     5372  0.002534  0.993789    448  0.001200\n",
       "18       0.001193   0.991712     5372  0.002982  0.988848    448  0.001200\n",
       "19       0.000887   0.992551     5372  0.002051  0.992217    448  0.001200\n",
       "20       0.000793   0.993223     5372  0.003178  0.989522    448  0.001200\n",
       "21       0.000874   0.992792     5372  0.001943  0.994239    448  0.001200\n",
       "22       0.004892   0.968810     5372  0.003582  0.987186    448  0.001200\n",
       "23       0.001487   0.988858     5372  0.002217  0.988982    448  0.000480\n",
       "24       0.000915   0.994382     5372  0.002288  0.990555    448  0.000480\n",
       "25       0.000790   0.994315     5372  0.002182  0.988567    448  0.000480\n",
       "26       0.000577   0.996377     5372  0.001856  0.990083    448  0.000480\n",
       "27       0.000577   0.996901     5372  0.002426  0.988589    448  0.000480\n",
       "28       0.000599   0.994965     5372  0.001855  0.991903    448  0.000480\n",
       "29       0.000496   0.996243     5372  0.002147  0.987186    448  0.000480\n",
       "30       0.000677   0.995590     5372  0.001686  0.992352    448  0.000480\n",
       "31       0.000545   0.995450     5372  0.001842  0.993639    448  0.000480\n",
       "32       0.000563   0.996313     5372  0.002000  0.995047    448  0.000480\n",
       "33       0.000526   0.996408     5372  0.002131  0.993699    448  0.000480\n",
       "34       0.000247   0.998544     5372  0.002118  0.993228    448  0.000192\n",
       "35       0.000384   0.997725     5372  0.001959  0.995047    448  0.000192\n",
       "36       0.000143   0.999669     5372  0.001535  0.994879    448  0.000192\n",
       "37       0.000423   0.996612     5372  0.001824  0.993228    448  0.000192\n",
       "38       0.000379   0.996707     5372  0.001912  0.993531    448  0.000192\n",
       "39       0.000332   0.996936     5372  0.001881  0.993531    448  0.000192\n",
       "40       0.000305   0.997826     5372  0.001928  0.995047    448  0.000192"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "trainer.train(40)\n",
    "trainer.complete_results"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cf9acdc1-3e51-4162-bd12-e44b9939765f",
   "metadata": {},
   "source": [
    "The model is significantly less accurate than the VGG16 model, but it has a gain in training and inference speed."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "1af84224-6797-4984-847a-8a594569cde3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       test_loss  test_acc  n_test        lr\n",
      "epoch                                       \n",
      "40      0.001972  0.989352   896.0  0.000192\n",
      "CPU times: user 182 ms, sys: 282 ms, total: 464 ms\n",
      "Wall time: 601 ms\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>test_loss</th>\n",
       "      <th>test_acc</th>\n",
       "      <th>n_test</th>\n",
       "      <th>lr</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>epoch</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>0.001972</td>\n",
       "      <td>0.989352</td>\n",
       "      <td>896.0</td>\n",
       "      <td>0.000192</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       test_loss  test_acc  n_test        lr\n",
       "epoch                                       \n",
       "40      0.001972  0.989352   896.0  0.000192"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "trainer.test()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "273c545b-4280-479b-b7db-a7250680600b",
   "metadata": {},
   "source": [
    "This model has a slight advantage in its training speed, but its accuracy is hindered by a proclivity for overfitting. When the raw data is available, it would make intuitive sense to apply a form of learning directly on the vectors instead of creating intermediate rasters. One problem lies in the expansion of the feature layer. 4 features * 10 strokes = 40 features! If there are any characters that share strokes in different positions, this model is forced to learn their effects independantly."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "38cc8348-de28-4451-80d6-f2731c0cc75c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1400x700 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, [ax1, ax2] = common_plot_setup(style=\"fivethirtyeight\", font=\"DejaVu Sans\", figsize=(14, 7), rows=1, cols=2)\n",
    "trainer.complete_results[[\"train_loss\", \"val_loss\"]].plot(ax=ax1, logy=True)\n",
    "trainer.complete_results[[\"train_acc\", \"val_acc\"]].plot(ax=ax2, logy=False)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ad851bc2-a5d1-49a1-a095-d28f1614eb33",
   "metadata": {},
   "source": [
    "We can also see where the mistakes came from. In this particular instance, the model had difficulty in distinguishing between ら and う. While the RNN offers greater precision than the CNN it the added freedom of possible values makes generalization more challenging."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "ffb96a32-8485-41f6-9f91-8379bfe43131",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Characters missed: 18\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>label</th>\n",
       "      <th>probability</th>\n",
       "      <th>char_id</th>\n",
       "      <th>character</th>\n",
       "      <th>romanization</th>\n",
       "      <th>stroke_count</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <th>label</th>\n",
       "      <th>char_id</th>\n",
       "      <th>character</th>\n",
       "      <th>romanization</th>\n",
       "      <th>indices</th>\n",
       "      <th>stroke_count</th>\n",
       "      <th>strokes_found</th>\n",
       "      <th>pred</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">5</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">hir_ra</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">ra</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">ら</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">ra</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">7, 8, 9, 10, 11</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">2</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">2.0</th>\n",
       "      <th>1</th>\n",
       "      <td>hir_u</td>\n",
       "      <td>0.663261</td>\n",
       "      <td>u</td>\n",
       "      <td>う</td>\n",
       "      <td>u</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>hir_ra</td>\n",
       "      <td>0.331278</td>\n",
       "      <td>ra</td>\n",
       "      <td>ら</td>\n",
       "      <td>ra</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hir_ti</td>\n",
       "      <td>0.004673</td>\n",
       "      <td>ti</td>\n",
       "      <td>ち</td>\n",
       "      <td>chi</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>hir_o</td>\n",
       "      <td>0.000154</td>\n",
       "      <td>o</td>\n",
       "      <td>お</td>\n",
       "      <td>o</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hir_vu</td>\n",
       "      <td>0.000150</td>\n",
       "      <td>vu</td>\n",
       "      <td>ゔ</td>\n",
       "      <td>vu</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">3</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">hir_u</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">u</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">う</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">u</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">14, 15, 16</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">2</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">2.0</th>\n",
       "      <th>1</th>\n",
       "      <td>hir_ra</td>\n",
       "      <td>0.787523</td>\n",
       "      <td>ra</td>\n",
       "      <td>ら</td>\n",
       "      <td>ra</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>hir_u</td>\n",
       "      <td>0.212055</td>\n",
       "      <td>u</td>\n",
       "      <td>う</td>\n",
       "      <td>u</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hir_vu</td>\n",
       "      <td>0.000220</td>\n",
       "      <td>vu</td>\n",
       "      <td>ゔ</td>\n",
       "      <td>vu</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>hir_dyu</td>\n",
       "      <td>0.000090</td>\n",
       "      <td>dyu</td>\n",
       "      <td>ぢゅ</td>\n",
       "      <td>dju</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hir_dya</td>\n",
       "      <td>0.000038</td>\n",
       "      <td>dya</td>\n",
       "      <td>ぢゃ</td>\n",
       "      <td>dja</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"10\" valign=\"top\">2</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">hir_e</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">e</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">え</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">e</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">1, 2</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">2</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">2.0</th>\n",
       "      <th>1</th>\n",
       "      <td>hir_e</td>\n",
       "      <td>0.410127</td>\n",
       "      <td>e</td>\n",
       "      <td>え</td>\n",
       "      <td>e</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>hir_ko</td>\n",
       "      <td>0.329411</td>\n",
       "      <td>ko</td>\n",
       "      <td>こ</td>\n",
       "      <td>ko</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hir_sa</td>\n",
       "      <td>0.250913</td>\n",
       "      <td>sa</td>\n",
       "      <td>さ</td>\n",
       "      <td>sa</td>\n",
       "      <td>2 or 3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>hir_ra</td>\n",
       "      <td>0.003271</td>\n",
       "      <td>ra</td>\n",
       "      <td>ら</td>\n",
       "      <td>ra</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hir_u</td>\n",
       "      <td>0.002212</td>\n",
       "      <td>u</td>\n",
       "      <td>う</td>\n",
       "      <td>u</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">hir_ko</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">ko</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">こ</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">ko</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">4, 5</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">2</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">2.0</th>\n",
       "      <th>1</th>\n",
       "      <td>hir_sa</td>\n",
       "      <td>0.479135</td>\n",
       "      <td>sa</td>\n",
       "      <td>さ</td>\n",
       "      <td>sa</td>\n",
       "      <td>2 or 3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>hir_go</td>\n",
       "      <td>0.342434</td>\n",
       "      <td>go</td>\n",
       "      <td>ご</td>\n",
       "      <td>go</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hir_do</td>\n",
       "      <td>0.059550</td>\n",
       "      <td>do</td>\n",
       "      <td>ど</td>\n",
       "      <td>do</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>hir_no</td>\n",
       "      <td>0.045421</td>\n",
       "      <td>no</td>\n",
       "      <td>の</td>\n",
       "      <td>no</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hir_tu</td>\n",
       "      <td>0.027484</td>\n",
       "      <td>tu</td>\n",
       "      <td>つ</td>\n",
       "      <td>tsu</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"30\" valign=\"top\">1</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">hir_dyo</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">dyo</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">ぢょ</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">djo</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">0</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">6</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">6.0</th>\n",
       "      <th>1</th>\n",
       "      <td>hir_da</td>\n",
       "      <td>0.997996</td>\n",
       "      <td>da</td>\n",
       "      <td>だ</td>\n",
       "      <td>da</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>hir_dyo</td>\n",
       "      <td>0.000850</td>\n",
       "      <td>dyo</td>\n",
       "      <td>ぢょ</td>\n",
       "      <td>djo</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hir_ze</td>\n",
       "      <td>0.000796</td>\n",
       "      <td>ze</td>\n",
       "      <td>ぜ</td>\n",
       "      <td>ze</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>hir_ga</td>\n",
       "      <td>0.000124</td>\n",
       "      <td>ga</td>\n",
       "      <td>が</td>\n",
       "      <td>ga</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hir_dya</td>\n",
       "      <td>0.000064</td>\n",
       "      <td>dya</td>\n",
       "      <td>ぢゃ</td>\n",
       "      <td>dja</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">hir_hya</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">hya</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">ひゃ</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">hya</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">3</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">4</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">4.0</th>\n",
       "      <th>1</th>\n",
       "      <td>hir_bi</td>\n",
       "      <td>0.799399</td>\n",
       "      <td>bi</td>\n",
       "      <td>び</td>\n",
       "      <td>bi</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>hir_hya</td>\n",
       "      <td>0.183084</td>\n",
       "      <td>hya</td>\n",
       "      <td>ひゃ</td>\n",
       "      <td>hya</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hir_zo</td>\n",
       "      <td>0.006690</td>\n",
       "      <td>zo</td>\n",
       "      <td>ぞ</td>\n",
       "      <td>zo</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>hir_byo</td>\n",
       "      <td>0.002524</td>\n",
       "      <td>byo</td>\n",
       "      <td>びょ</td>\n",
       "      <td>byo</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hir_gu</td>\n",
       "      <td>0.002132</td>\n",
       "      <td>gu</td>\n",
       "      <td>ぐ</td>\n",
       "      <td>gu</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">hir_mya</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">mya</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">みゃ</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">mya</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">6</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">5</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">5.0</th>\n",
       "      <th>1</th>\n",
       "      <td>hir_rya</td>\n",
       "      <td>0.794063</td>\n",
       "      <td>rya</td>\n",
       "      <td>りゃ</td>\n",
       "      <td>rya</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>hir_mya</td>\n",
       "      <td>0.090865</td>\n",
       "      <td>mya</td>\n",
       "      <td>みゃ</td>\n",
       "      <td>mya</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hir_myu</td>\n",
       "      <td>0.065101</td>\n",
       "      <td>myu</td>\n",
       "      <td>みゅ</td>\n",
       "      <td>myu</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>hir_nyu</td>\n",
       "      <td>0.033875</td>\n",
       "      <td>nyu</td>\n",
       "      <td>にゅ</td>\n",
       "      <td>nyu</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hir_tya</td>\n",
       "      <td>0.012276</td>\n",
       "      <td>tya</td>\n",
       "      <td>ちゃ</td>\n",
       "      <td>cha</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">hir_si</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">si</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">し</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">shi</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">12</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">1</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">1.0</th>\n",
       "      <th>1</th>\n",
       "      <td>hir_hi</td>\n",
       "      <td>0.283329</td>\n",
       "      <td>hi</td>\n",
       "      <td>ひ</td>\n",
       "      <td>hi</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>hir_no</td>\n",
       "      <td>0.206594</td>\n",
       "      <td>no</td>\n",
       "      <td>の</td>\n",
       "      <td>no</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hir_si</td>\n",
       "      <td>0.118848</td>\n",
       "      <td>si</td>\n",
       "      <td>し</td>\n",
       "      <td>shi</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>hir_u</td>\n",
       "      <td>0.081252</td>\n",
       "      <td>u</td>\n",
       "      <td>う</td>\n",
       "      <td>u</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hir_ya</td>\n",
       "      <td>0.072286</td>\n",
       "      <td>ya</td>\n",
       "      <td>や</td>\n",
       "      <td>ya</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">hir_syo</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">syo</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">しょ</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">sho</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">13</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">3</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">3.0</th>\n",
       "      <th>1</th>\n",
       "      <td>hir_hyo</td>\n",
       "      <td>0.777883</td>\n",
       "      <td>hyo</td>\n",
       "      <td>ひょ</td>\n",
       "      <td>hyo</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>hir_ha</td>\n",
       "      <td>0.186435</td>\n",
       "      <td>ha</td>\n",
       "      <td>は</td>\n",
       "      <td>ha</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hir_po</td>\n",
       "      <td>0.018310</td>\n",
       "      <td>po</td>\n",
       "      <td>ぽ</td>\n",
       "      <td>po</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>hir_pyo</td>\n",
       "      <td>0.008720</td>\n",
       "      <td>pyo</td>\n",
       "      <td>ぴょ</td>\n",
       "      <td>pyo</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hir_syo</td>\n",
       "      <td>0.005176</td>\n",
       "      <td>syo</td>\n",
       "      <td>しょ</td>\n",
       "      <td>sho</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">hir_wo</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">wo</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">を</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">wo</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">17</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">3</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">3.0</th>\n",
       "      <th>1</th>\n",
       "      <td>hir_a</td>\n",
       "      <td>0.998027</td>\n",
       "      <td>a</td>\n",
       "      <td>あ</td>\n",
       "      <td>a</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>hir_wo</td>\n",
       "      <td>0.001497</td>\n",
       "      <td>wo</td>\n",
       "      <td>を</td>\n",
       "      <td>wo</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hir_ta</td>\n",
       "      <td>0.000339</td>\n",
       "      <td>ta</td>\n",
       "      <td>た</td>\n",
       "      <td>ta</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>hir_tyu</td>\n",
       "      <td>0.000050</td>\n",
       "      <td>tyu</td>\n",
       "      <td>ちゅ</td>\n",
       "      <td>chu</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hir_da</td>\n",
       "      <td>0.000031</td>\n",
       "      <td>da</td>\n",
       "      <td>だ</td>\n",
       "      <td>da</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                                                label  \\\n",
       "count label   char_id character romanization indices         stroke_count strokes_found pred            \n",
       "5     hir_ra  ra      ら         ra           7, 8, 9, 10, 11 2            2.0           1       hir_u   \n",
       "                                                                                        2      hir_ra   \n",
       "                                                                                        3      hir_ti   \n",
       "                                                                                        4       hir_o   \n",
       "                                                                                        5      hir_vu   \n",
       "3     hir_u   u       う         u            14, 15, 16      2            2.0           1      hir_ra   \n",
       "                                                                                        2       hir_u   \n",
       "                                                                                        3      hir_vu   \n",
       "                                                                                        4     hir_dyu   \n",
       "                                                                                        5     hir_dya   \n",
       "2     hir_e   e       え         e            1, 2            2            2.0           1       hir_e   \n",
       "                                                                                        2      hir_ko   \n",
       "                                                                                        3      hir_sa   \n",
       "                                                                                        4      hir_ra   \n",
       "                                                                                        5       hir_u   \n",
       "      hir_ko  ko      こ         ko           4, 5            2            2.0           1      hir_sa   \n",
       "                                                                                        2      hir_go   \n",
       "                                                                                        3      hir_do   \n",
       "                                                                                        4      hir_no   \n",
       "                                                                                        5      hir_tu   \n",
       "1     hir_dyo dyo     ぢょ        djo          0               6            6.0           1      hir_da   \n",
       "                                                                                        2     hir_dyo   \n",
       "                                                                                        3      hir_ze   \n",
       "                                                                                        4      hir_ga   \n",
       "                                                                                        5     hir_dya   \n",
       "      hir_hya hya     ひゃ        hya          3               4            4.0           1      hir_bi   \n",
       "                                                                                        2     hir_hya   \n",
       "                                                                                        3      hir_zo   \n",
       "                                                                                        4     hir_byo   \n",
       "                                                                                        5      hir_gu   \n",
       "      hir_mya mya     みゃ        mya          6               5            5.0           1     hir_rya   \n",
       "                                                                                        2     hir_mya   \n",
       "                                                                                        3     hir_myu   \n",
       "                                                                                        4     hir_nyu   \n",
       "                                                                                        5     hir_tya   \n",
       "      hir_si  si      し         shi          12              1            1.0           1      hir_hi   \n",
       "                                                                                        2      hir_no   \n",
       "                                                                                        3      hir_si   \n",
       "                                                                                        4       hir_u   \n",
       "                                                                                        5      hir_ya   \n",
       "      hir_syo syo     しょ        sho          13              3            3.0           1     hir_hyo   \n",
       "                                                                                        2      hir_ha   \n",
       "                                                                                        3      hir_po   \n",
       "                                                                                        4     hir_pyo   \n",
       "                                                                                        5     hir_syo   \n",
       "      hir_wo  wo      を         wo           17              3            3.0           1       hir_a   \n",
       "                                                                                        2      hir_wo   \n",
       "                                                                                        3      hir_ta   \n",
       "                                                                                        4     hir_tyu   \n",
       "                                                                                        5      hir_da   \n",
       "\n",
       "                                                                                              probability  \\\n",
       "count label   char_id character romanization indices         stroke_count strokes_found pred                \n",
       "5     hir_ra  ra      ら         ra           7, 8, 9, 10, 11 2            2.0           1        0.663261   \n",
       "                                                                                        2        0.331278   \n",
       "                                                                                        3        0.004673   \n",
       "                                                                                        4        0.000154   \n",
       "                                                                                        5        0.000150   \n",
       "3     hir_u   u       う         u            14, 15, 16      2            2.0           1        0.787523   \n",
       "                                                                                        2        0.212055   \n",
       "                                                                                        3        0.000220   \n",
       "                                                                                        4        0.000090   \n",
       "                                                                                        5        0.000038   \n",
       "2     hir_e   e       え         e            1, 2            2            2.0           1        0.410127   \n",
       "                                                                                        2        0.329411   \n",
       "                                                                                        3        0.250913   \n",
       "                                                                                        4        0.003271   \n",
       "                                                                                        5        0.002212   \n",
       "      hir_ko  ko      こ         ko           4, 5            2            2.0           1        0.479135   \n",
       "                                                                                        2        0.342434   \n",
       "                                                                                        3        0.059550   \n",
       "                                                                                        4        0.045421   \n",
       "                                                                                        5        0.027484   \n",
       "1     hir_dyo dyo     ぢょ        djo          0               6            6.0           1        0.997996   \n",
       "                                                                                        2        0.000850   \n",
       "                                                                                        3        0.000796   \n",
       "                                                                                        4        0.000124   \n",
       "                                                                                        5        0.000064   \n",
       "      hir_hya hya     ひゃ        hya          3               4            4.0           1        0.799399   \n",
       "                                                                                        2        0.183084   \n",
       "                                                                                        3        0.006690   \n",
       "                                                                                        4        0.002524   \n",
       "                                                                                        5        0.002132   \n",
       "      hir_mya mya     みゃ        mya          6               5            5.0           1        0.794063   \n",
       "                                                                                        2        0.090865   \n",
       "                                                                                        3        0.065101   \n",
       "                                                                                        4        0.033875   \n",
       "                                                                                        5        0.012276   \n",
       "      hir_si  si      し         shi          12              1            1.0           1        0.283329   \n",
       "                                                                                        2        0.206594   \n",
       "                                                                                        3        0.118848   \n",
       "                                                                                        4        0.081252   \n",
       "                                                                                        5        0.072286   \n",
       "      hir_syo syo     しょ        sho          13              3            3.0           1        0.777883   \n",
       "                                                                                        2        0.186435   \n",
       "                                                                                        3        0.018310   \n",
       "                                                                                        4        0.008720   \n",
       "                                                                                        5        0.005176   \n",
       "      hir_wo  wo      を         wo           17              3            3.0           1        0.998027   \n",
       "                                                                                        2        0.001497   \n",
       "                                                                                        3        0.000339   \n",
       "                                                                                        4        0.000050   \n",
       "                                                                                        5        0.000031   \n",
       "\n",
       "                                                                                             char_id  \\\n",
       "count label   char_id character romanization indices         stroke_count strokes_found pred           \n",
       "5     hir_ra  ra      ら         ra           7, 8, 9, 10, 11 2            2.0           1          u   \n",
       "                                                                                        2         ra   \n",
       "                                                                                        3         ti   \n",
       "                                                                                        4          o   \n",
       "                                                                                        5         vu   \n",
       "3     hir_u   u       う         u            14, 15, 16      2            2.0           1         ra   \n",
       "                                                                                        2          u   \n",
       "                                                                                        3         vu   \n",
       "                                                                                        4        dyu   \n",
       "                                                                                        5        dya   \n",
       "2     hir_e   e       え         e            1, 2            2            2.0           1          e   \n",
       "                                                                                        2         ko   \n",
       "                                                                                        3         sa   \n",
       "                                                                                        4         ra   \n",
       "                                                                                        5          u   \n",
       "      hir_ko  ko      こ         ko           4, 5            2            2.0           1         sa   \n",
       "                                                                                        2         go   \n",
       "                                                                                        3         do   \n",
       "                                                                                        4         no   \n",
       "                                                                                        5         tu   \n",
       "1     hir_dyo dyo     ぢょ        djo          0               6            6.0           1         da   \n",
       "                                                                                        2        dyo   \n",
       "                                                                                        3         ze   \n",
       "                                                                                        4         ga   \n",
       "                                                                                        5        dya   \n",
       "      hir_hya hya     ひゃ        hya          3               4            4.0           1         bi   \n",
       "                                                                                        2        hya   \n",
       "                                                                                        3         zo   \n",
       "                                                                                        4        byo   \n",
       "                                                                                        5         gu   \n",
       "      hir_mya mya     みゃ        mya          6               5            5.0           1        rya   \n",
       "                                                                                        2        mya   \n",
       "                                                                                        3        myu   \n",
       "                                                                                        4        nyu   \n",
       "                                                                                        5        tya   \n",
       "      hir_si  si      し         shi          12              1            1.0           1         hi   \n",
       "                                                                                        2         no   \n",
       "                                                                                        3         si   \n",
       "                                                                                        4          u   \n",
       "                                                                                        5         ya   \n",
       "      hir_syo syo     しょ        sho          13              3            3.0           1        hyo   \n",
       "                                                                                        2         ha   \n",
       "                                                                                        3         po   \n",
       "                                                                                        4        pyo   \n",
       "                                                                                        5        syo   \n",
       "      hir_wo  wo      を         wo           17              3            3.0           1          a   \n",
       "                                                                                        2         wo   \n",
       "                                                                                        3         ta   \n",
       "                                                                                        4        tyu   \n",
       "                                                                                        5         da   \n",
       "\n",
       "                                                                                             character  \\\n",
       "count label   char_id character romanization indices         stroke_count strokes_found pred             \n",
       "5     hir_ra  ra      ら         ra           7, 8, 9, 10, 11 2            2.0           1            う   \n",
       "                                                                                        2            ら   \n",
       "                                                                                        3            ち   \n",
       "                                                                                        4            お   \n",
       "                                                                                        5            ゔ   \n",
       "3     hir_u   u       う         u            14, 15, 16      2            2.0           1            ら   \n",
       "                                                                                        2            う   \n",
       "                                                                                        3            ゔ   \n",
       "                                                                                        4           ぢゅ   \n",
       "                                                                                        5           ぢゃ   \n",
       "2     hir_e   e       え         e            1, 2            2            2.0           1            え   \n",
       "                                                                                        2            こ   \n",
       "                                                                                        3            さ   \n",
       "                                                                                        4            ら   \n",
       "                                                                                        5            う   \n",
       "      hir_ko  ko      こ         ko           4, 5            2            2.0           1            さ   \n",
       "                                                                                        2            ご   \n",
       "                                                                                        3            ど   \n",
       "                                                                                        4            の   \n",
       "                                                                                        5            つ   \n",
       "1     hir_dyo dyo     ぢょ        djo          0               6            6.0           1            だ   \n",
       "                                                                                        2           ぢょ   \n",
       "                                                                                        3            ぜ   \n",
       "                                                                                        4            が   \n",
       "                                                                                        5           ぢゃ   \n",
       "      hir_hya hya     ひゃ        hya          3               4            4.0           1            び   \n",
       "                                                                                        2           ひゃ   \n",
       "                                                                                        3            ぞ   \n",
       "                                                                                        4           びょ   \n",
       "                                                                                        5            ぐ   \n",
       "      hir_mya mya     みゃ        mya          6               5            5.0           1           りゃ   \n",
       "                                                                                        2           みゃ   \n",
       "                                                                                        3           みゅ   \n",
       "                                                                                        4           にゅ   \n",
       "                                                                                        5           ちゃ   \n",
       "      hir_si  si      し         shi          12              1            1.0           1            ひ   \n",
       "                                                                                        2            の   \n",
       "                                                                                        3            し   \n",
       "                                                                                        4            う   \n",
       "                                                                                        5            や   \n",
       "      hir_syo syo     しょ        sho          13              3            3.0           1           ひょ   \n",
       "                                                                                        2            は   \n",
       "                                                                                        3            ぽ   \n",
       "                                                                                        4           ぴょ   \n",
       "                                                                                        5           しょ   \n",
       "      hir_wo  wo      を         wo           17              3            3.0           1            あ   \n",
       "                                                                                        2            を   \n",
       "                                                                                        3            た   \n",
       "                                                                                        4           ちゅ   \n",
       "                                                                                        5            だ   \n",
       "\n",
       "                                                                                             romanization  \\\n",
       "count label   char_id character romanization indices         stroke_count strokes_found pred                \n",
       "5     hir_ra  ra      ら         ra           7, 8, 9, 10, 11 2            2.0           1               u   \n",
       "                                                                                        2              ra   \n",
       "                                                                                        3             chi   \n",
       "                                                                                        4               o   \n",
       "                                                                                        5              vu   \n",
       "3     hir_u   u       う         u            14, 15, 16      2            2.0           1              ra   \n",
       "                                                                                        2               u   \n",
       "                                                                                        3              vu   \n",
       "                                                                                        4             dju   \n",
       "                                                                                        5             dja   \n",
       "2     hir_e   e       え         e            1, 2            2            2.0           1               e   \n",
       "                                                                                        2              ko   \n",
       "                                                                                        3              sa   \n",
       "                                                                                        4              ra   \n",
       "                                                                                        5               u   \n",
       "      hir_ko  ko      こ         ko           4, 5            2            2.0           1              sa   \n",
       "                                                                                        2              go   \n",
       "                                                                                        3              do   \n",
       "                                                                                        4              no   \n",
       "                                                                                        5             tsu   \n",
       "1     hir_dyo dyo     ぢょ        djo          0               6            6.0           1              da   \n",
       "                                                                                        2             djo   \n",
       "                                                                                        3              ze   \n",
       "                                                                                        4              ga   \n",
       "                                                                                        5             dja   \n",
       "      hir_hya hya     ひゃ        hya          3               4            4.0           1              bi   \n",
       "                                                                                        2             hya   \n",
       "                                                                                        3              zo   \n",
       "                                                                                        4             byo   \n",
       "                                                                                        5              gu   \n",
       "      hir_mya mya     みゃ        mya          6               5            5.0           1             rya   \n",
       "                                                                                        2             mya   \n",
       "                                                                                        3             myu   \n",
       "                                                                                        4             nyu   \n",
       "                                                                                        5             cha   \n",
       "      hir_si  si      し         shi          12              1            1.0           1              hi   \n",
       "                                                                                        2              no   \n",
       "                                                                                        3             shi   \n",
       "                                                                                        4               u   \n",
       "                                                                                        5              ya   \n",
       "      hir_syo syo     しょ        sho          13              3            3.0           1             hyo   \n",
       "                                                                                        2              ha   \n",
       "                                                                                        3              po   \n",
       "                                                                                        4             pyo   \n",
       "                                                                                        5             sho   \n",
       "      hir_wo  wo      を         wo           17              3            3.0           1               a   \n",
       "                                                                                        2              wo   \n",
       "                                                                                        3              ta   \n",
       "                                                                                        4             chu   \n",
       "                                                                                        5              da   \n",
       "\n",
       "                                                                                             stroke_count  \n",
       "count label   char_id character romanization indices         stroke_count strokes_found pred               \n",
       "5     hir_ra  ra      ら         ra           7, 8, 9, 10, 11 2            2.0           1               2  \n",
       "                                                                                        2               2  \n",
       "                                                                                        3               2  \n",
       "                                                                                        4               3  \n",
       "                                                                                        5               4  \n",
       "3     hir_u   u       う         u            14, 15, 16      2            2.0           1               2  \n",
       "                                                                                        2               2  \n",
       "                                                                                        3               4  \n",
       "                                                                                        4               6  \n",
       "                                                                                        5               7  \n",
       "2     hir_e   e       え         e            1, 2            2            2.0           1               2  \n",
       "                                                                                        2               2  \n",
       "                                                                                        3          2 or 3  \n",
       "                                                                                        4               2  \n",
       "                                                                                        5               2  \n",
       "      hir_ko  ko      こ         ko           4, 5            2            2.0           1          2 or 3  \n",
       "                                                                                        2               4  \n",
       "                                                                                        3               4  \n",
       "                                                                                        4               1  \n",
       "                                                                                        5               1  \n",
       "1     hir_dyo dyo     ぢょ        djo          0               6            6.0           1               6  \n",
       "                                                                                        2               6  \n",
       "                                                                                        3               5  \n",
       "                                                                                        4               5  \n",
       "                                                                                        5               7  \n",
       "      hir_hya hya     ひゃ        hya          3               4            4.0           1               3  \n",
       "                                                                                        2               4  \n",
       "                                                                                        3               3  \n",
       "                                                                                        4               5  \n",
       "                                                                                        5               3  \n",
       "      hir_mya mya     みゃ        mya          6               5            5.0           1               5  \n",
       "                                                                                        2               5  \n",
       "                                                                                        3               4  \n",
       "                                                                                        4               5  \n",
       "                                                                                        5               5  \n",
       "      hir_si  si      し         shi          12              1            1.0           1               1  \n",
       "                                                                                        2               1  \n",
       "                                                                                        3               1  \n",
       "                                                                                        4               2  \n",
       "                                                                                        5               3  \n",
       "      hir_syo syo     しょ        sho          13              3            3.0           1               3  \n",
       "                                                                                        2               3  \n",
       "                                                                                        3               5  \n",
       "                                                                                        4               4  \n",
       "                                                                                        5               3  \n",
       "      hir_wo  wo      を         wo           17              3            3.0           1               3  \n",
       "                                                                                        2               3  \n",
       "                                                                                        3               4  \n",
       "                                                                                        4               4  \n",
       "                                                                                        5               6  "
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "misses = IncorrectCharacters(trainer, transformed_dataset, data_paths.get_classes())\n",
    "print(\"Characters missed:\", len(misses))\n",
    "misses.overview(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1e0939ca-741d-4368-be57-abf1eda937fd",
   "metadata": {},
   "source": [
    "## Model 3: LSTM + GRU"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4964b993-8e60-47b6-85b1-d570e69bf671",
   "metadata": {},
   "source": [
    "The next model presents a unique solution to the idea of repeated strokes. Because virtually identical strokes may between several strokes in different positions the model would benefit from a component that extracts information from them independantly.\n",
    "\n",
    "Rather than use a supervised approach that requires additional training to label and train against commonly occuring shapes and radicals, it is more efficient to use the hidden outputs directly as inputs into a second RNN.\n",
    "\n",
    "This can be achieved by exploding the strokes into the batch dimension of the first RNN and immediately reshaping the hidden outputs to the original batch and stroke sizes.\n",
    "\n",
    "When the strokes exist as additional outputs for individual evaluation, the first RNN can generalize the transformation of a stroke regardless of its position but feed the strokes sequentially into a second RNN. This becomes (batch, stroke, points, features) -> (batches and strokes, points, features) for RNN1, while RNN2 takes in (batches, strokes, hidden layers) for it's prediction.\n",
    "\n",
    "To implement this, I chose an LSTM for the initial layer since the memory of LSTMs provides a distinct advantage when the amount of timesteps is high, but for the second RNN I chose a GRU for faster performance given that the amount of strokes is much smaller in number."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "c5e5e0fa-3880-4913-a314-4d62ad6cd2e4",
   "metadata": {},
   "outputs": [],
   "source": [
    "class CharacterIdentifier(nn.Module):\n",
    "    def __init__(self, lstm_features, lstm_hidden, gru_hidden, num_classes):\n",
    "        super().__init__()\n",
    "        self._lstm_hidden = lstm_hidden\n",
    "        self._gru_hidden = gru_hidden\n",
    "        self.lstm = nn.LSTM(lstm_features, lstm_hidden, num_layers=1, batch_first=True)\n",
    "        self.gru = nn.GRU(lstm_hidden, gru_hidden, num_layers=1, batch_first=True)\n",
    "        self.linear = nn.Linear(gru_hidden, num_classes)\n",
    "        self.linear_alt = nn.Linear(lstm_hidden, num_classes)\n",
    "        self.dropout = nn.Dropout(p = 0.1)\n",
    "        self.lstm_h = (torch.zeros((0)), torch.zeros((0)))\n",
    "        self.gru_h = torch.zeros((0))\n",
    "        self.dummy_param = nn.Parameter(torch.empty(0))\n",
    "\n",
    "    def _init_hidden_state(self, previous, x, hidden_size):\n",
    "        if tuple(previous.shape) != (1, x.shape[0], hidden_size):\n",
    "            device = self.dummy_param.device\n",
    "            return torch.zeros(1, x.shape[0], hidden_size).to(device)\n",
    "        return previous\n",
    "\n",
    "    def _init_lstm_hidden_states(self, x):\n",
    "        h = self._init_hidden_state(self.lstm_h[0], x, self._lstm_hidden)\n",
    "        if self.lstm_h[0] is not h:\n",
    "            self.lstm_h = (h, h.clone())\n",
    "\n",
    "    def _lstm_forward(self, x):\n",
    "        # Strokes will be converted to batches since this LSTM only needs to learn how to read a shape from a sequence of points.\n",
    "        batches, strokes, sequence, features = x.shape\n",
    "        x = x.view(batches * strokes, sequence, features)\n",
    "        self._init_lstm_hidden_states(x)\n",
    "        _, (h, c) = self.lstm(x, self.lstm_h)\n",
    "        out = h.view(batches, strokes, self._lstm_hidden)\n",
    "        return out, (h, c)\n",
    "\n",
    "    def _gru_forward(self, x):\n",
    "        self.gru_h = self._init_hidden_state(self.gru_h, x, self._gru_hidden)\n",
    "        _, h_g = self.gru(x, self.gru_h)\n",
    "        out = h_g.view(-1, self._gru_hidden)\n",
    "        return out, h_g\n",
    "\n",
    "    def forward(self, inputs):\n",
    "        out, (h_l, c_l) = self._lstm_forward(inputs)\n",
    "        out, h_g = self._gru_forward(out)\n",
    "        out = self.dropout(out)\n",
    "        out = self.linear(out)\n",
    "        return out"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d47f727b-8f2b-41aa-bfae-cb1f68b13952",
   "metadata": {},
   "source": [
    "The data inputs and transformations will be nearly identical here, but in this case there is no longer a need to melt strokes into the feature dimension. The model object handles the reshaping internally to be passed into the LSTM and subsequent GRU.\n",
    "\n",
    "Unlike the previous model, during training I also noticed that the random gaussian noise tranformation, did not carry a significant positive benefit. This may be due to the fact that when melting the strokes into the batches, the first lstm effictively has an enlarged dataset with a smaller risk of overfitting."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "0e64e7a7-9996-43ed-b603-e0601e21855f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'train': <torch.utils.data.dataloader.DataLoader at 0x7ffaef5390d0>,\n",
       " 'val': <torch.utils.data.dataloader.DataLoader at 0x7ffb13674d50>,\n",
       " 'test': <torch.utils.data.dataloader.DataLoader at 0x7ffb1349e650>}"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_transforms = [ToBSplines(120), ArrayToTensor(),\n",
    "                   ExtractAngles(include_magnitudes=False),\n",
    "                   InputMinMaxTransformer(None, old_min, old_max, new_min = -1),\n",
    "                   InputRecenter(None, std_shift),\n",
    "                  ]\n",
    "# During testing, the gaussian noise presented no tangible benefits to this model and will be skipped.\n",
    "train_transforms = transforms.Compose(test_transforms + [EmptyStrokePadder(stroke_count)])\n",
    "test_transforms = transforms.Compose(test_transforms + [EmptyStrokePadder(stroke_count)])\n",
    "transformed_dataset = StrokeDataset(data_paths.get_data(), transform=test_transforms)\n",
    "split_keys = {\"train\": train_transforms, \"val\": test_transforms, \"test\": test_transforms}\n",
    "split_datapaths = dict(zip(split_keys, data_paths.train_val_test_split(train_size=.7, test_size=.2, random_state=SEED)))\n",
    "datasets = {key: StrokeDataset(split_datapaths[key], transform=value) for key, value in split_keys.items()}\n",
    "dataloaders = {key: torch.utils.data.DataLoader(datasets[key], batch_size=batch_size, shuffle=True, num_workers=5, drop_last=True) for key in split_keys.keys()}\n",
    "dataloaders"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "7546341d-e817-457d-b4c0-b4e704832088",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "108\n"
     ]
    }
   ],
   "source": [
    "lstm_features = 4\n",
    "lstm_hidden_size = 20\n",
    "gru_hidden_size = 100\n",
    "learning_rate=0.004\n",
    "n_classes = np.unique(full_dataset.labels).size\n",
    "print(n_classes)\n",
    "\n",
    "model = CharacterIdentifier(lstm_features, lstm_hidden_size, gru_hidden_size, n_classes)\n",
    "criterion = nn.CrossEntropyLoss()\n",
    "optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)\n",
    "scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=11, gamma=0.4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "87a581e1-e539-4a41-8c66-1c335c23567a",
   "metadata": {},
   "outputs": [],
   "source": [
    "trainer = CharacterTrainer(model, optimizer, criterion, dataloaders, scheduler=scheduler, single_input=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "07666be1-6a8d-4e4e-9c3f-08a6f4e4b900",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "1        0.177243   0.115485   5372.0  0.113975  0.270455  448.0  0.004\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "2        0.082312    0.49944   5372.0  0.053163  0.672727  448.0  0.004\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "3        0.044366   0.723694   5372.0  0.032735  0.779545  448.0  0.004\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "4        0.029577    0.80541   5372.0  0.025744  0.829545  448.0  0.004\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "5        0.023321   0.842164   5372.0  0.018366  0.870455  448.0  0.004\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "6        0.019028   0.869963   5372.0  0.019135  0.861364  448.0  0.004\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "7         0.01763   0.878918   5372.0  0.013597  0.909091  448.0  0.004\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val     lr\n",
      "epoch                                                                 \n",
      "8        0.016983   0.882463   5372.0  0.019387    0.875  448.0  0.004\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "9        0.016144   0.888246   5372.0  0.017726  0.877273  448.0  0.004\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "10       0.015309   0.883769   5372.0  0.012993  0.918182  448.0  0.004\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "11       0.011794   0.909888   5372.0  0.011095  0.918182  448.0  0.004\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "12       0.008264   0.932649   5372.0  0.009455  0.929545  448.0  0.0016\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "13       0.006878   0.943843   5372.0  0.010971  0.959091  448.0  0.0016\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "14       0.006136   0.949067   5372.0  0.007294  0.940909  448.0  0.0016\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "15        0.00605   0.951866   5372.0  0.010383  0.931818  448.0  0.0016\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "16       0.007526    0.94459   5372.0   0.00766  0.952273  448.0  0.0016\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "17       0.005904   0.950933   5372.0   0.00914  0.929545  448.0  0.0016\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "18       0.004732   0.961754   5372.0  0.007223  0.956818  448.0  0.0016\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "19       0.004598   0.965485   5372.0  0.006566  0.972727  448.0  0.0016\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val      lr\n",
      "epoch                                                                  \n",
      "20       0.004976   0.961381   5372.0  0.005754     0.95  448.0  0.0016\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "21        0.00427   0.965299   5372.0  0.006957  0.947727  448.0  0.0016\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "22       0.003967   0.967537   5372.0  0.008149  0.947727  448.0  0.0016\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "23       0.003185   0.976306   5372.0  0.005397  0.972727  448.0  0.00064\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "24       0.002521    0.98209   5372.0  0.006575  0.968182  448.0  0.00064\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val       lr\n",
      "epoch                                                                   \n",
      "25       0.002618   0.980037   5372.0  0.005009    0.975  448.0  0.00064\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "26       0.002328   0.981716   5372.0  0.005576  0.977273  448.0  0.00064\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "27       0.002138   0.982649   5372.0  0.005368  0.968182  448.0  0.00064\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "28       0.002115   0.984142   5372.0  0.005764  0.972727  448.0  0.00064\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "29       0.001953   0.986194   5372.0  0.005512  0.972727  448.0  0.00064\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "30       0.001922   0.983955   5372.0  0.005097  0.977273  448.0  0.00064\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "31       0.001682    0.98694   5372.0  0.005204  0.979545  448.0  0.00064\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "32       0.001913   0.987313   5372.0  0.005988  0.972727  448.0  0.00064\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
      "epoch                                                                    \n",
      "33       0.002165   0.983209   5372.0  0.005084  0.972727  448.0  0.00064\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val        lr\n",
      "epoch                                                                     \n",
      "34       0.001461   0.988993   5372.0  0.004244  0.979545  448.0  0.000256\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val        lr\n",
      "epoch                                                                     \n",
      "35       0.001365   0.990485   5372.0  0.005144  0.972727  448.0  0.000256\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val        lr\n",
      "epoch                                                                     \n",
      "36       0.001228   0.991604   5372.0  0.003819  0.977273  448.0  0.000256\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val        lr\n",
      "epoch                                                                     \n",
      "37       0.001036    0.99347   5372.0  0.003808  0.984091  448.0  0.000256\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val        lr\n",
      "epoch                                                                    \n",
      "38       0.001068   0.993284   5372.0   0.00451    0.975  448.0  0.000256\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val        lr\n",
      "epoch                                                                     \n",
      "39       0.001024   0.992724   5372.0  0.005193  0.970455  448.0  0.000256\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val        lr\n",
      "epoch                                                                    \n",
      "40       0.000977   0.994216   5372.0  0.004104    0.975  448.0  0.000256\n",
      "CPU times: user 34.9 s, sys: 23.5 s, total: 58.3 s\n",
      "Wall time: 1min 25s\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>train_loss</th>\n",
       "      <th>train_acc</th>\n",
       "      <th>n_train</th>\n",
       "      <th>val_loss</th>\n",
       "      <th>val_acc</th>\n",
       "      <th>n_val</th>\n",
       "      <th>lr</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>epoch</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.177243</td>\n",
       "      <td>0.115485</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.113975</td>\n",
       "      <td>0.270455</td>\n",
       "      <td>448</td>\n",
       "      <td>0.004000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.082312</td>\n",
       "      <td>0.499440</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.053163</td>\n",
       "      <td>0.672727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.004000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.044366</td>\n",
       "      <td>0.723694</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.032735</td>\n",
       "      <td>0.779545</td>\n",
       "      <td>448</td>\n",
       "      <td>0.004000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.029577</td>\n",
       "      <td>0.805410</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.025744</td>\n",
       "      <td>0.829545</td>\n",
       "      <td>448</td>\n",
       "      <td>0.004000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.023321</td>\n",
       "      <td>0.842164</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.018366</td>\n",
       "      <td>0.870455</td>\n",
       "      <td>448</td>\n",
       "      <td>0.004000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.019028</td>\n",
       "      <td>0.869963</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.019135</td>\n",
       "      <td>0.861364</td>\n",
       "      <td>448</td>\n",
       "      <td>0.004000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.017630</td>\n",
       "      <td>0.878918</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.013597</td>\n",
       "      <td>0.909091</td>\n",
       "      <td>448</td>\n",
       "      <td>0.004000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0.016983</td>\n",
       "      <td>0.882463</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.019387</td>\n",
       "      <td>0.875000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.004000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.016144</td>\n",
       "      <td>0.888246</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.017726</td>\n",
       "      <td>0.877273</td>\n",
       "      <td>448</td>\n",
       "      <td>0.004000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.015309</td>\n",
       "      <td>0.883769</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.012993</td>\n",
       "      <td>0.918182</td>\n",
       "      <td>448</td>\n",
       "      <td>0.004000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0.011794</td>\n",
       "      <td>0.909888</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.011095</td>\n",
       "      <td>0.918182</td>\n",
       "      <td>448</td>\n",
       "      <td>0.004000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>0.008264</td>\n",
       "      <td>0.932649</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.009455</td>\n",
       "      <td>0.929545</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>0.006878</td>\n",
       "      <td>0.943843</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.010971</td>\n",
       "      <td>0.959091</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>0.006136</td>\n",
       "      <td>0.949067</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.007294</td>\n",
       "      <td>0.940909</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0.006050</td>\n",
       "      <td>0.951866</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.010383</td>\n",
       "      <td>0.931818</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>0.007526</td>\n",
       "      <td>0.944590</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.007660</td>\n",
       "      <td>0.952273</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>0.005904</td>\n",
       "      <td>0.950933</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.009140</td>\n",
       "      <td>0.929545</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>0.004732</td>\n",
       "      <td>0.961754</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.007223</td>\n",
       "      <td>0.956818</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>0.004598</td>\n",
       "      <td>0.965485</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.006566</td>\n",
       "      <td>0.972727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>0.004976</td>\n",
       "      <td>0.961381</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.005754</td>\n",
       "      <td>0.950000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>0.004270</td>\n",
       "      <td>0.965299</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.006957</td>\n",
       "      <td>0.947727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>0.003967</td>\n",
       "      <td>0.967537</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.008149</td>\n",
       "      <td>0.947727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.001600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>0.003185</td>\n",
       "      <td>0.976306</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.005397</td>\n",
       "      <td>0.972727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000640</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>0.002521</td>\n",
       "      <td>0.982090</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.006575</td>\n",
       "      <td>0.968182</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000640</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>0.002618</td>\n",
       "      <td>0.980037</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.005009</td>\n",
       "      <td>0.975000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000640</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>0.002328</td>\n",
       "      <td>0.981716</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.005576</td>\n",
       "      <td>0.977273</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000640</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>0.002138</td>\n",
       "      <td>0.982649</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.005368</td>\n",
       "      <td>0.968182</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000640</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>0.002115</td>\n",
       "      <td>0.984142</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.005764</td>\n",
       "      <td>0.972727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000640</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>0.001953</td>\n",
       "      <td>0.986194</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.005512</td>\n",
       "      <td>0.972727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000640</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>0.001922</td>\n",
       "      <td>0.983955</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.005097</td>\n",
       "      <td>0.977273</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000640</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>0.001682</td>\n",
       "      <td>0.986940</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.005204</td>\n",
       "      <td>0.979545</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000640</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>0.001913</td>\n",
       "      <td>0.987313</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.005988</td>\n",
       "      <td>0.972727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000640</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>0.002165</td>\n",
       "      <td>0.983209</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.005084</td>\n",
       "      <td>0.972727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000640</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>0.001461</td>\n",
       "      <td>0.988993</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.004244</td>\n",
       "      <td>0.979545</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000256</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>0.001365</td>\n",
       "      <td>0.990485</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.005144</td>\n",
       "      <td>0.972727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000256</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>0.001228</td>\n",
       "      <td>0.991604</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.003819</td>\n",
       "      <td>0.977273</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000256</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>0.001036</td>\n",
       "      <td>0.993470</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.003808</td>\n",
       "      <td>0.984091</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000256</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>0.001068</td>\n",
       "      <td>0.993284</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.004510</td>\n",
       "      <td>0.975000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000256</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>0.001024</td>\n",
       "      <td>0.992724</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.005193</td>\n",
       "      <td>0.970455</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000256</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>0.000977</td>\n",
       "      <td>0.994216</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.004104</td>\n",
       "      <td>0.975000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.000256</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       train_loss  train_acc  n_train  val_loss   val_acc  n_val        lr\n",
       "epoch                                                                     \n",
       "1        0.177243   0.115485     5372  0.113975  0.270455    448  0.004000\n",
       "2        0.082312   0.499440     5372  0.053163  0.672727    448  0.004000\n",
       "3        0.044366   0.723694     5372  0.032735  0.779545    448  0.004000\n",
       "4        0.029577   0.805410     5372  0.025744  0.829545    448  0.004000\n",
       "5        0.023321   0.842164     5372  0.018366  0.870455    448  0.004000\n",
       "6        0.019028   0.869963     5372  0.019135  0.861364    448  0.004000\n",
       "7        0.017630   0.878918     5372  0.013597  0.909091    448  0.004000\n",
       "8        0.016983   0.882463     5372  0.019387  0.875000    448  0.004000\n",
       "9        0.016144   0.888246     5372  0.017726  0.877273    448  0.004000\n",
       "10       0.015309   0.883769     5372  0.012993  0.918182    448  0.004000\n",
       "11       0.011794   0.909888     5372  0.011095  0.918182    448  0.004000\n",
       "12       0.008264   0.932649     5372  0.009455  0.929545    448  0.001600\n",
       "13       0.006878   0.943843     5372  0.010971  0.959091    448  0.001600\n",
       "14       0.006136   0.949067     5372  0.007294  0.940909    448  0.001600\n",
       "15       0.006050   0.951866     5372  0.010383  0.931818    448  0.001600\n",
       "16       0.007526   0.944590     5372  0.007660  0.952273    448  0.001600\n",
       "17       0.005904   0.950933     5372  0.009140  0.929545    448  0.001600\n",
       "18       0.004732   0.961754     5372  0.007223  0.956818    448  0.001600\n",
       "19       0.004598   0.965485     5372  0.006566  0.972727    448  0.001600\n",
       "20       0.004976   0.961381     5372  0.005754  0.950000    448  0.001600\n",
       "21       0.004270   0.965299     5372  0.006957  0.947727    448  0.001600\n",
       "22       0.003967   0.967537     5372  0.008149  0.947727    448  0.001600\n",
       "23       0.003185   0.976306     5372  0.005397  0.972727    448  0.000640\n",
       "24       0.002521   0.982090     5372  0.006575  0.968182    448  0.000640\n",
       "25       0.002618   0.980037     5372  0.005009  0.975000    448  0.000640\n",
       "26       0.002328   0.981716     5372  0.005576  0.977273    448  0.000640\n",
       "27       0.002138   0.982649     5372  0.005368  0.968182    448  0.000640\n",
       "28       0.002115   0.984142     5372  0.005764  0.972727    448  0.000640\n",
       "29       0.001953   0.986194     5372  0.005512  0.972727    448  0.000640\n",
       "30       0.001922   0.983955     5372  0.005097  0.977273    448  0.000640\n",
       "31       0.001682   0.986940     5372  0.005204  0.979545    448  0.000640\n",
       "32       0.001913   0.987313     5372  0.005988  0.972727    448  0.000640\n",
       "33       0.002165   0.983209     5372  0.005084  0.972727    448  0.000640\n",
       "34       0.001461   0.988993     5372  0.004244  0.979545    448  0.000256\n",
       "35       0.001365   0.990485     5372  0.005144  0.972727    448  0.000256\n",
       "36       0.001228   0.991604     5372  0.003819  0.977273    448  0.000256\n",
       "37       0.001036   0.993470     5372  0.003808  0.984091    448  0.000256\n",
       "38       0.001068   0.993284     5372  0.004510  0.975000    448  0.000256\n",
       "39       0.001024   0.992724     5372  0.005193  0.970455    448  0.000256\n",
       "40       0.000977   0.994216     5372  0.004104  0.975000    448  0.000256"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "trainer.train(40)\n",
    "trainer.complete_results"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bc91fe77-c402-4f85-83e1-eacc3aafe88d",
   "metadata": {},
   "source": [
    "This architecture has a small immediate training/inference speed boost, but also contains the potential for an even larger one in a scenario where real time on screen identification with live feedback is desired. Instead of passing the entire tensor of inputs each time a new stroke is added, the model states could be cached with strokes being passed in in one at a time."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "3c331223-7709-42ca-a53e-fcef43f2b24e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       test_loss  test_acc  n_test        lr\n",
      "epoch                                       \n",
      "40      0.007008  0.968182   896.0  0.000256\n",
      "CPU times: user 87.2 ms, sys: 265 ms, total: 353 ms\n",
      "Wall time: 585 ms\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>test_loss</th>\n",
       "      <th>test_acc</th>\n",
       "      <th>n_test</th>\n",
       "      <th>lr</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>epoch</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>0.007008</td>\n",
       "      <td>0.968182</td>\n",
       "      <td>896.0</td>\n",
       "      <td>0.000256</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       test_loss  test_acc  n_test        lr\n",
       "epoch                                       \n",
       "40      0.007008  0.968182   896.0  0.000256"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "trainer.test()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d8826f7f-01e4-4c19-b908-a2381e5d8eea",
   "metadata": {},
   "source": [
    "While the reduced risk of overfitting allowed the RNN to continue improving for 40 epochs without the addition of the random gaussian noise, the model performs significantly worse in its predictions. One advantage is that despite having two RNNs instead of one the reduced neurons within cut the training time for the same amount of epochs in half. There could be merit in increasing the amount of hidden outputs from the LSTM to better inform the GRU, but an even more interesting possibility would be to make the underlying LSTM inputs more generalizable while passing additional features directly into the GRU."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "id": "4bb73aab-fed6-45c1-b09c-71afcddf2a50",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1400x700 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, [ax1, ax2] = common_plot_setup(style=\"fivethirtyeight\", font=\"DejaVu Sans\", figsize=(14, 7), rows=1, cols=2)\n",
    "trainer.complete_results[[\"train_loss\", \"val_loss\"]].plot(ax=ax1, logy=True)\n",
    "trainer.complete_results[[\"train_acc\", \"val_acc\"]].plot(ax=ax2, logy=False)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3ecead99-845f-419f-a93f-0e079e880507",
   "metadata": {},
   "source": [
    "In the case of this model, the misclassifications happen much more often and there is a stronger bias that hinders the prediction of certain characters (especially う). There is one small advantage in that the count of strokes appears to be more important in the predicitions that are made."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "id": "dcee3e1f-9940-491e-8ace-e97a5978caf1",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Characters missed: 68\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>label</th>\n",
       "      <th>probability</th>\n",
       "      <th>char_id</th>\n",
       "      <th>character</th>\n",
       "      <th>romanization</th>\n",
       "      <th>stroke_count</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <th>label</th>\n",
       "      <th>char_id</th>\n",
       "      <th>character</th>\n",
       "      <th>romanization</th>\n",
       "      <th>indices</th>\n",
       "      <th>stroke_count</th>\n",
       "      <th>strokes_found</th>\n",
       "      <th>pred</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">14</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">hir_tu</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">tu</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">つ</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">tsu</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">1</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">1.0</th>\n",
       "      <th>1</th>\n",
       "      <td>hir_ro</td>\n",
       "      <td>0.703673</td>\n",
       "      <td>ro</td>\n",
       "      <td>ろ</td>\n",
       "      <td>ro</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>hir_tu</td>\n",
       "      <td>0.267252</td>\n",
       "      <td>tu</td>\n",
       "      <td>つ</td>\n",
       "      <td>tsu</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hir_no</td>\n",
       "      <td>0.027508</td>\n",
       "      <td>no</td>\n",
       "      <td>の</td>\n",
       "      <td>no</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>hir_vu</td>\n",
       "      <td>0.000507</td>\n",
       "      <td>vu</td>\n",
       "      <td>ゔ</td>\n",
       "      <td>vu</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hir_ra</td>\n",
       "      <td>0.000272</td>\n",
       "      <td>ra</td>\n",
       "      <td>ら</td>\n",
       "      <td>ra</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <th>...</th>\n",
       "      <th>...</th>\n",
       "      <th>...</th>\n",
       "      <th>...</th>\n",
       "      <th>...</th>\n",
       "      <th>...</th>\n",
       "      <th>...</th>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">1</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">hir_wo</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">wo</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">を</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">wo</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">60</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">3</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">3.0</th>\n",
       "      <th>1</th>\n",
       "      <td>hir_se</td>\n",
       "      <td>0.995950</td>\n",
       "      <td>se</td>\n",
       "      <td>せ</td>\n",
       "      <td>se</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>hir_wo</td>\n",
       "      <td>0.002502</td>\n",
       "      <td>wo</td>\n",
       "      <td>を</td>\n",
       "      <td>wo</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hir_ti</td>\n",
       "      <td>0.000569</td>\n",
       "      <td>ti</td>\n",
       "      <td>ち</td>\n",
       "      <td>chi</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>hir_o</td>\n",
       "      <td>0.000268</td>\n",
       "      <td>o</td>\n",
       "      <td>お</td>\n",
       "      <td>o</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hir_he</td>\n",
       "      <td>0.000247</td>\n",
       "      <td>he</td>\n",
       "      <td>へ</td>\n",
       "      <td>he</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>140 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                                                                                 label  \\\n",
       "count label  char_id character romanization indices                                            stroke_count strokes_found pred           \n",
       "14    hir_tu tu      つ         tsu          40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51,... 1            1.0           1     hir_ro   \n",
       "                                                                                                                          2     hir_tu   \n",
       "                                                                                                                          3     hir_no   \n",
       "                                                                                                                          4     hir_vu   \n",
       "                                                                                                                          5     hir_ra   \n",
       "...                                                                                                                                ...   \n",
       "1     hir_wo wo      を         wo           60                                                 3            3.0           1     hir_se   \n",
       "                                                                                                                          2     hir_wo   \n",
       "                                                                                                                          3     hir_ti   \n",
       "                                                                                                                          4      hir_o   \n",
       "                                                                                                                          5     hir_he   \n",
       "\n",
       "                                                                                                                                probability  \\\n",
       "count label  char_id character romanization indices                                            stroke_count strokes_found pred                \n",
       "14    hir_tu tu      つ         tsu          40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51,... 1            1.0           1        0.703673   \n",
       "                                                                                                                          2        0.267252   \n",
       "                                                                                                                          3        0.027508   \n",
       "                                                                                                                          4        0.000507   \n",
       "                                                                                                                          5        0.000272   \n",
       "...                                                                                                                                     ...   \n",
       "1     hir_wo wo      を         wo           60                                                 3            3.0           1        0.995950   \n",
       "                                                                                                                          2        0.002502   \n",
       "                                                                                                                          3        0.000569   \n",
       "                                                                                                                          4        0.000268   \n",
       "                                                                                                                          5        0.000247   \n",
       "\n",
       "                                                                                                                               char_id  \\\n",
       "count label  char_id character romanization indices                                            stroke_count strokes_found pred           \n",
       "14    hir_tu tu      つ         tsu          40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51,... 1            1.0           1         ro   \n",
       "                                                                                                                          2         tu   \n",
       "                                                                                                                          3         no   \n",
       "                                                                                                                          4         vu   \n",
       "                                                                                                                          5         ra   \n",
       "...                                                                                                                                ...   \n",
       "1     hir_wo wo      を         wo           60                                                 3            3.0           1         se   \n",
       "                                                                                                                          2         wo   \n",
       "                                                                                                                          3         ti   \n",
       "                                                                                                                          4          o   \n",
       "                                                                                                                          5         he   \n",
       "\n",
       "                                                                                                                               character  \\\n",
       "count label  char_id character romanization indices                                            stroke_count strokes_found pred             \n",
       "14    hir_tu tu      つ         tsu          40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51,... 1            1.0           1            ろ   \n",
       "                                                                                                                          2            つ   \n",
       "                                                                                                                          3            の   \n",
       "                                                                                                                          4            ゔ   \n",
       "                                                                                                                          5            ら   \n",
       "...                                                                                                                                  ...   \n",
       "1     hir_wo wo      を         wo           60                                                 3            3.0           1            せ   \n",
       "                                                                                                                          2            を   \n",
       "                                                                                                                          3            ち   \n",
       "                                                                                                                          4            お   \n",
       "                                                                                                                          5            へ   \n",
       "\n",
       "                                                                                                                               romanization  \\\n",
       "count label  char_id character romanization indices                                            stroke_count strokes_found pred                \n",
       "14    hir_tu tu      つ         tsu          40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51,... 1            1.0           1              ro   \n",
       "                                                                                                                          2             tsu   \n",
       "                                                                                                                          3              no   \n",
       "                                                                                                                          4              vu   \n",
       "                                                                                                                          5              ra   \n",
       "...                                                                                                                                     ...   \n",
       "1     hir_wo wo      を         wo           60                                                 3            3.0           1              se   \n",
       "                                                                                                                          2              wo   \n",
       "                                                                                                                          3             chi   \n",
       "                                                                                                                          4               o   \n",
       "                                                                                                                          5              he   \n",
       "\n",
       "                                                                                                                               stroke_count  \n",
       "count label  char_id character romanization indices                                            stroke_count strokes_found pred               \n",
       "14    hir_tu tu      つ         tsu          40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51,... 1            1.0           1               1  \n",
       "                                                                                                                          2               1  \n",
       "                                                                                                                          3               1  \n",
       "                                                                                                                          4               4  \n",
       "                                                                                                                          5               2  \n",
       "...                                                                                                                                     ...  \n",
       "1     hir_wo wo      を         wo           60                                                 3            3.0           1               3  \n",
       "                                                                                                                          2               3  \n",
       "                                                                                                                          3               2  \n",
       "                                                                                                                          4               3  \n",
       "                                                                                                                          5               1  \n",
       "\n",
       "[140 rows x 6 columns]"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "misses = IncorrectCharacters(trainer, transformed_dataset, data_paths.get_classes())\n",
    "print(\"Characters missed:\", len(misses))\n",
    "misses.overview(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "feef765c-edff-4f7e-9f69-87b31d7eac94",
   "metadata": {},
   "source": [
    "## Model 4: LSTM + GRU with relative/absolute feature seperation"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e65a91ba-8652-4ca3-bc25-6d2ccc1cf24f",
   "metadata": {},
   "source": [
    "The last model takes the general idea of model 3 and expands on it further. Instead of passing the raw sequential points and angles into the LSTM, data aggregations are fed directly into the GRU, while more generic centered and scaled inputs are fed into the LSTM.\n",
    "\n",
    "The data is extracted to form two tensors with the following inputs:\n",
    "- GRU (using the aggregated strokes as timesteps) inputs:\n",
    "  - Mean of all X positions\n",
    "  - Mean of all y positions\n",
    "  - Standard deviation of all X positions (proxy for stroke length/width since points are sequential)\n",
    "  - Standard deviation of all y positions (proxy for stroke length/width since points are sequential)\n",
    "  - Difference between each first x coordinate in the sequence and the mean (additional proxy for length, width and direction).\n",
    "  - Difference between each first y coordinate in the sequence and the mean (additional proxy for length, width and direction).\n",
    "  - Difference between each last x coordinate in the sequence and the mean (additional proxy for length, width and direction).\n",
    "  - Difference between each last y coordinate in the sequence and the mean (additional proxy for length, width and direction).\n",
    "  - Mean of all sine values between points and their distances (component to measure angular strength continuously)\n",
    "  - Mean of all cosine values between points and their distances (component to measure angular strength continuously)\n",
    "  - LSTM hidden layers (the remaining relative data to inform the shape)\n",
    "- LSTM (using the relative point level data) inputs\n",
    "  - Normalized sine values after centering against their mean.\n",
    "  - Normalized cosine values after centering against their mean.\n",
    "  - Normalized differencing of sine values (measures angular change and can be represented in a single column of values while maintaining its continous nature)\n",
    "  - Normalized length between points (helps quantify importance between angles)\n",
    " \n",
    "One advantage of this idea is clearly demonstrated when considering a straight line. A whole sequence of drastically different values could be passed into the input of the LSTM for simple reasons such as translation, differences in scale, and rotation. By extracting the aggregate absolute data for direct input into the GRU, a generalization of a given sequence of points is more likely to occur.\n",
    "\n",
    "Because the inputs will be changing drastically, new aggregated values must be extracted for the purpose of scaling and normalization."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "id": "271e7177-1cb2-4dd2-9cc6-d144fe67b142",
   "metadata": {},
   "outputs": [],
   "source": [
    "aggregator = StrokeDatasetAggregator(data_paths, transforms.Compose([ToBSplines(120), # In this model all LSTM inputs are derived from differenced point data,\n",
    "                                                                     ArrayToTensor(),\n",
    "                                                                     StrokeExtractAbsolute()]), 1 # Class that divides the input tensor into the two sets.\n",
    "                                    )"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3d1c621b-bdae-40df-a519-d2b14ae37c8f",
   "metadata": {},
   "source": [
    "Taking the means and standard deviations of points will effectively shift the range of values that was previously -1 to 1. The following code will first create the GRU inputs for the entire dataset and then aggregate values for the purpose of centering and scaling the results before being input into the model.\n",
    "\n",
    "The relative values fed into the LSTM will be skipped and instead be normalized in respect to each observation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "6443e671-ab1d-492a-9d70-1dacb47b00b0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min: tensor([[-0.8473, -0.6155,  0.0015,  0.0000, -0.9135, -0.5204, -0.6727, -0.9282,\n",
      "         -0.6701, -0.9993]])\n",
      "Max: tensor([[0.9361, 0.8746, 0.5997, 0.5980, 0.5269, 1.0623, 1.0578, 0.3485, 0.9999,\n",
      "         0.2845]])\n"
     ]
    }
   ],
   "source": [
    "old_min, old_max = aggregator.get_min_max(dim=0, keepdim=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "712495ca-f36d-41ba-8d21-dd63d73a1cca",
   "metadata": {},
   "source": [
    "As expected, the aggregate values have lower ranges than when points were evaluated individually. Once again, the rotational data contains a strong skew from the effect of proper stroke order and will be recentered. In addition recentering is important for the standard deviations since the skew of the values on a -1 to 1 scale will heavily favor the negative since horizontal and vertical strokes with almost no standard deviation along an axis are common and large spacing between points is very uncommon.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "c14f4a3b-5da6-469c-bb0c-a81f121fcdbc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean: tensor([[ 0.0877,  0.1766,  0.1527,  0.1812, -0.1649,  0.2838,  0.1776, -0.2088,\n",
      "          0.4164, -0.4831]])\n",
      "STD: tensor([[0.3846, 0.2809, 0.1202, 0.1570, 0.2173, 0.2877, 0.2460, 0.2223, 0.3681,\n",
      "         0.3501]])\n"
     ]
    }
   ],
   "source": [
    "means, _ = aggregator.get_mean_std(dim=0, keepdim=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a05b1dc4-4953-4858-8d7c-797169e78438",
   "metadata": {},
   "source": [
    "As with the other model, the recentering will be skipped on the features directly related to the absolute x and y coordinates."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "5f418c09-63e3-44cc-b922-19de0ee4f314",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([[ 0.0485,  0.0631, -0.4945, -0.3941,  0.0395,  0.0163, -0.0174,  0.1269,\n",
      "          0.3012, -0.1958]])\n",
      "tensor([[ 0.0000,  0.0000, -0.4945, -0.3941,  0.0395,  0.0163, -0.0174,  0.1269,\n",
      "          0.3012, -0.1958]])\n"
     ]
    }
   ],
   "source": [
    "std_means = (means - old_min) / (old_max - old_min) * 2 - 1\n",
    "print(std_means)\n",
    "std_shift = f.pad(std_means[:,2:], (2, 0))\n",
    "print(std_shift)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9cd573ba-d363-4d6b-a095-38b9c4a2306f",
   "metadata": {},
   "source": [
    "The custom classes for this model are defined."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "id": "3bfc88a1-2ed3-451d-b5cf-de8550d69010",
   "metadata": {},
   "outputs": [],
   "source": [
    "class ConformedStrokeDataset(StrokeDataset):\n",
    "    \"\"\"Child class of the Stroke Dataset to handle the seperate LSTM and GRU model inputs and apply 0 value padding to the strokes after all other transformations.\"\"\"\n",
    "    def __init__(self, path_df, transform, stroke_count):\n",
    "        super().__init__(path_df, transform)\n",
    "        self._padder = EmptyStrokePadder(stroke_count)\n",
    "\n",
    "    def __getitem__(self, idx):\n",
    "        (relative, absolute), label = super().__getitem__(idx)\n",
    "        return self._padder(relative), self._padder(absolute), label\n",
    "\n",
    "class ExtractedCharacterIdentifier(nn.Module):\n",
    "    def __init__(self, lstm_features, lstm_hidden, gru_additional_features, gru_hidden, num_classes):\n",
    "        super().__init__()\n",
    "        self._lstm_hidden = lstm_hidden\n",
    "        self._gru_hidden = gru_hidden\n",
    "        self._gru_features = gru_additional_features\n",
    "        self.lstm = nn.LSTM(lstm_features, lstm_hidden, num_layers=1, batch_first=True)\n",
    "        self.gru = nn.GRU(lstm_hidden+gru_additional_features, gru_hidden, num_layers=1, batch_first=True)\n",
    "        self.linear = nn.Linear(gru_hidden, num_classes)\n",
    "        self.linear_first = nn.Linear(lstm_hidden, lstm_hidden)\n",
    "        self.dropout = nn.Dropout(p = 0.2)\n",
    "        self.lstm_h = (torch.zeros((0)), torch.zeros((0)))\n",
    "        self.gru_h = torch.zeros((0))\n",
    "        self.dummy_param = nn.Parameter(torch.empty(0))\n",
    "        self.mish = nn.Mish()\n",
    "\n",
    "    def _init_hidden_state(self, previous, x, hidden_size):\n",
    "        if tuple(previous.shape) != (1, x.shape[0], hidden_size):\n",
    "            device = self.dummy_param.device\n",
    "            return torch.zeros(1, x.shape[0], hidden_size).to(device)\n",
    "        return previous\n",
    "\n",
    "    def _init_lstm_hidden_states(self, x):\n",
    "        h = self._init_hidden_state(self.lstm_h[0], x, self._lstm_hidden)\n",
    "        if self.lstm_h[0] is not h:\n",
    "            self.lstm_h = (h, h.clone())\n",
    "\n",
    "    def _lstm_forward(self, x):\n",
    "        # Strokes will be converted to batches since this LSTM only needs to learn how to read a shape from a sequence of points.\n",
    "        batches, strokes, sequence, features = x.shape\n",
    "        x = x.view(batches * strokes, sequence, features)\n",
    "        self._init_lstm_hidden_states(x)\n",
    "        _, (h, c) = self.lstm(x, self.lstm_h)\n",
    "        out = h.view(batches, strokes, self._lstm_hidden)\n",
    "        return out, (h, c)\n",
    "\n",
    "    def _gru_forward(self, x):\n",
    "        self.gru_h = self._init_hidden_state(self.gru_h, x, self._gru_hidden)\n",
    "        _, h_g = self.gru(x, self.gru_h)\n",
    "        out = h_g.view(-1, self._gru_hidden)\n",
    "        return out, h_g\n",
    "\n",
    "    def forward(self, inputs):\n",
    "        lstm_vals, gru_vals = inputs\n",
    "        out, (h_l, c_l) = self._lstm_forward(lstm_vals)\n",
    "        out = torch.cat([gru_vals, out], dim=-1)\n",
    "        out, h_g = self._gru_forward(out)\n",
    "        out = self.dropout(out)\n",
    "        out = self.linear(out)\n",
    "        return out"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ec2991b6-40f5-4484-a4db-6c12bf95b40e",
   "metadata": {},
   "source": [
    "In addition to the seperation of features, A small amount of gaussian noise will be added to both inputs. It is stronger on the inputs to the GRU layer since the effective batches for it remain the same. For much of those parameters, the effect can almost be compared to the translation, scaling, and rotation image augmentation techniques of the CNN, but affecting each stroke independently."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "id": "4adb863c-b06d-4c82-8ab3-fe9eeac471ea",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'train': <torch.utils.data.dataloader.DataLoader at 0x7ffb0bcfc3d0>,\n",
       " 'val': <torch.utils.data.dataloader.DataLoader at 0x7ffb0bcfcf10>,\n",
       " 'test': <torch.utils.data.dataloader.DataLoader at 0x7ffb0bcfc050>}"
      ]
     },
     "execution_count": 86,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_transforms = [ToBSplines(120), ArrayToTensor(), StrokeExtractAbsolute(),\n",
    "                   InputNormalizer(0,1), # Normalizes the already relative LSTM inputs for each feature.\n",
    "                   InputMinMaxTransformer(1, old_min, old_max, new_min = -1), # Normalizes the GRU inputs from -1 to 1 (without adjusting the angle parameters)\n",
    "                   InputRecenter(1, std_shift) # Recenters the standard deviation features of the GRU inputs to the means within the entire dataset.\n",
    "                  ]\n",
    "# Training dataset will be augmented by random transformations.\n",
    "train_transforms = transforms.Compose(test_transforms + [InputGaussianNoise(0, .04), InputGaussianNoise(1, .1)])\n",
    "#train_transforms = transforms.Compose(test_transforms +  [InputGaussianNoise(1, .1)])\n",
    "test_transforms = transforms.Compose(test_transforms)\n",
    "transformed_dataset = StrokeDataset(data_paths.get_data(), transform=test_transforms)\n",
    "conformed_dataset = ConformedStrokeDataset(data_paths.get_data(), transform=test_transforms, stroke_count=stroke_count)\n",
    "split_keys = {\"train\": train_transforms, \"val\": test_transforms, \"test\": test_transforms}\n",
    "split_datapaths = dict(zip(split_keys, data_paths.train_val_test_split(train_size=.7, test_size=.2, random_state=SEED)))\n",
    "datasets = {key: ConformedStrokeDataset(split_datapaths[key], transform=value, stroke_count=stroke_count) for key, value in split_keys.items()}\n",
    "dataloaders = {key: torch.utils.data.DataLoader(datasets[key], batch_size=batch_size, shuffle=True, num_workers=5, drop_last=True) for key in split_keys.keys()}\n",
    "dataloaders"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "id": "d4d55efe-91e8-42ba-b111-15eb14cabbd4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "108\n"
     ]
    }
   ],
   "source": [
    "lstm_features = 4\n",
    "gru_features = 10\n",
    "lstm_hidden_size = 10\n",
    "gru_hidden_size = 100\n",
    "learning_rate=0.005\n",
    "n_classes = np.unique(full_dataset.labels).size\n",
    "print(n_classes)\n",
    "\n",
    "model = ExtractedCharacterIdentifier(lstm_features, lstm_hidden_size, gru_features, gru_hidden_size, n_classes)\n",
    "criterion = nn.CrossEntropyLoss()\n",
    "optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)\n",
    "scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=11, gamma=0.4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "id": "8eeea46f-14ba-45e6-a33c-a0d1a6bc05e6",
   "metadata": {},
   "outputs": [],
   "source": [
    "trainer = CharacterTrainer(model, optimizer, criterion, dataloaders, scheduler=scheduler, single_input=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "id": "e31372b7-f1d6-4e2b-a796-1097a18e28c7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "1         0.13963   0.270336   5372.0  0.055159  0.606818  448.0  0.005\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "2        0.038713   0.742724   5372.0  0.016581  0.895455  448.0  0.005\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "3        0.018557   0.883396   5372.0  0.008394  0.945455  448.0  0.005\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "4        0.010794   0.923507   5372.0  0.007102  0.936364  448.0  0.005\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "5        0.008575   0.943284   5372.0  0.007405  0.940909  448.0  0.005\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "6        0.008059   0.948321   5372.0  0.003769  0.968182  448.0  0.005\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "7        0.006086   0.960261   5372.0  0.001893  0.988636  448.0  0.005\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "8        0.005367   0.966604   5372.0    0.0029  0.984091  448.0  0.005\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "9        0.003831   0.979104   5372.0  0.000903  0.990909  448.0  0.005\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "10       0.004822   0.972201   5372.0  0.003475  0.979545  448.0  0.005\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "11        0.00431   0.977239   5372.0   0.00104  0.995455  448.0  0.005\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "12       0.001392   0.992164   5372.0  0.000287  0.997727  448.0  0.002\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val     lr\n",
      "epoch                                                                 \n",
      "13       0.001008   0.993657   5372.0  0.000182      1.0  448.0  0.002\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "14       0.001024   0.994216   5372.0  0.001981  0.990909  448.0  0.002\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "15       0.001141    0.99459   5372.0  0.000243  0.997727  448.0  0.002\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val     lr\n",
      "epoch                                                                 \n",
      "16       0.001066   0.994403   5372.0  0.000198      1.0  448.0  0.002\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val     lr\n",
      "epoch                                                                 \n",
      "17       0.000646   0.996642   5372.0  0.000148      1.0  448.0  0.002\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "18       0.000872   0.995709   5372.0  0.002752  0.979545  448.0  0.002\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "19       0.001856   0.989552   5372.0  0.002104  0.986364  448.0  0.002\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val     lr\n",
      "epoch                                                                  \n",
      "20       0.000949    0.99459   5372.0  0.000401  0.997727  448.0  0.002\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val     lr\n",
      "epoch                                                                 \n",
      "21       0.001109   0.992724   5372.0  0.000084      1.0  448.0  0.002\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val     lr\n",
      "epoch                                                                 \n",
      "22        0.00078   0.996269   5372.0  0.000117      1.0  448.0  0.002\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "23       0.000548   0.997388   5372.0  0.000151  0.997727  448.0  0.0008\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "24       0.000388   0.998134   5372.0  0.000259  0.997727  448.0  0.0008\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val      lr\n",
      "epoch                                                                  \n",
      "25       0.000371   0.997388   5372.0  0.000033      1.0  448.0  0.0008\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val      lr\n",
      "epoch                                                                  \n",
      "26       0.000491   0.997201   5372.0  0.000047      1.0  448.0  0.0008\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val      lr\n",
      "epoch                                                                  \n",
      "27       0.000317   0.998507   5372.0   0.00005      1.0  448.0  0.0008\n",
      "       train_loss  train_acc  n_train  val_loss   val_acc  n_val      lr\n",
      "epoch                                                                   \n",
      "28       0.000257   0.998881   5372.0  0.000327  0.997727  448.0  0.0008\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val      lr\n",
      "epoch                                                                  \n",
      "29       0.000304   0.997761   5372.0   0.00003      1.0  448.0  0.0008\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val      lr\n",
      "epoch                                                                  \n",
      "30       0.000374   0.997761   5372.0  0.000024      1.0  448.0  0.0008\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val      lr\n",
      "epoch                                                                  \n",
      "31       0.000371   0.997761   5372.0  0.000085      1.0  448.0  0.0008\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val      lr\n",
      "epoch                                                                  \n",
      "32       0.000452   0.997575   5372.0  0.000048      1.0  448.0  0.0008\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val      lr\n",
      "epoch                                                                  \n",
      "33       0.000218   0.998694   5372.0  0.000011      1.0  448.0  0.0008\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val       lr\n",
      "epoch                                                                   \n",
      "34        0.00013   0.999254   5372.0   0.00001      1.0  448.0  0.00032\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val       lr\n",
      "epoch                                                                   \n",
      "35       0.000176   0.999254   5372.0  0.000017      1.0  448.0  0.00032\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val       lr\n",
      "epoch                                                                   \n",
      "36         0.0002   0.998694   5372.0  0.000062      1.0  448.0  0.00032\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val       lr\n",
      "epoch                                                                   \n",
      "37       0.000201   0.998694   5372.0  0.000009      1.0  448.0  0.00032\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val       lr\n",
      "epoch                                                                   \n",
      "38       0.000187   0.998881   5372.0  0.000008      1.0  448.0  0.00032\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val       lr\n",
      "epoch                                                                   \n",
      "39       0.000188   0.998881   5372.0  0.000053      1.0  448.0  0.00032\n",
      "       train_loss  train_acc  n_train  val_loss  val_acc  n_val       lr\n",
      "epoch                                                                   \n",
      "40       0.000284   0.998321   5372.0  0.000009      1.0  448.0  0.00032\n",
      "CPU times: user 36.8 s, sys: 23.6 s, total: 1min\n",
      "Wall time: 1min 40s\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>train_loss</th>\n",
       "      <th>train_acc</th>\n",
       "      <th>n_train</th>\n",
       "      <th>val_loss</th>\n",
       "      <th>val_acc</th>\n",
       "      <th>n_val</th>\n",
       "      <th>lr</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>epoch</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.139630</td>\n",
       "      <td>0.270336</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.055159</td>\n",
       "      <td>0.606818</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.038713</td>\n",
       "      <td>0.742724</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.016581</td>\n",
       "      <td>0.895455</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.018557</td>\n",
       "      <td>0.883396</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.008394</td>\n",
       "      <td>0.945455</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.010794</td>\n",
       "      <td>0.923507</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.007102</td>\n",
       "      <td>0.936364</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.008575</td>\n",
       "      <td>0.943284</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.007405</td>\n",
       "      <td>0.940909</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.008059</td>\n",
       "      <td>0.948321</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.003769</td>\n",
       "      <td>0.968182</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.006086</td>\n",
       "      <td>0.960261</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.001893</td>\n",
       "      <td>0.988636</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0.005367</td>\n",
       "      <td>0.966604</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.002900</td>\n",
       "      <td>0.984091</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.003831</td>\n",
       "      <td>0.979104</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000903</td>\n",
       "      <td>0.990909</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.004822</td>\n",
       "      <td>0.972201</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.003475</td>\n",
       "      <td>0.979545</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0.004310</td>\n",
       "      <td>0.977239</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.001040</td>\n",
       "      <td>0.995455</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>0.001392</td>\n",
       "      <td>0.992164</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000287</td>\n",
       "      <td>0.997727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>0.001008</td>\n",
       "      <td>0.993657</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000182</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>0.001024</td>\n",
       "      <td>0.994216</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.001981</td>\n",
       "      <td>0.990909</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0.001141</td>\n",
       "      <td>0.994590</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000243</td>\n",
       "      <td>0.997727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>0.001066</td>\n",
       "      <td>0.994403</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000198</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>0.000646</td>\n",
       "      <td>0.996642</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000148</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>0.000872</td>\n",
       "      <td>0.995709</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.002752</td>\n",
       "      <td>0.979545</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>0.001856</td>\n",
       "      <td>0.989552</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.002104</td>\n",
       "      <td>0.986364</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>0.000949</td>\n",
       "      <td>0.994590</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000401</td>\n",
       "      <td>0.997727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>0.001109</td>\n",
       "      <td>0.992724</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000084</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>0.000780</td>\n",
       "      <td>0.996269</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000117</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>0.000548</td>\n",
       "      <td>0.997388</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000151</td>\n",
       "      <td>0.997727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00080</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>0.000388</td>\n",
       "      <td>0.998134</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000259</td>\n",
       "      <td>0.997727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00080</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>0.000371</td>\n",
       "      <td>0.997388</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000033</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00080</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>0.000491</td>\n",
       "      <td>0.997201</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000047</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00080</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>0.000317</td>\n",
       "      <td>0.998507</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000050</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00080</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>0.000257</td>\n",
       "      <td>0.998881</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000327</td>\n",
       "      <td>0.997727</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00080</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>0.000304</td>\n",
       "      <td>0.997761</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000030</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00080</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>0.000374</td>\n",
       "      <td>0.997761</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000024</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00080</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>0.000371</td>\n",
       "      <td>0.997761</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000085</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00080</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>0.000452</td>\n",
       "      <td>0.997575</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000048</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00080</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>0.000218</td>\n",
       "      <td>0.998694</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000011</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00080</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>0.000130</td>\n",
       "      <td>0.999254</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000010</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00032</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>0.000176</td>\n",
       "      <td>0.999254</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000017</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00032</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>0.000200</td>\n",
       "      <td>0.998694</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000062</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00032</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>0.000201</td>\n",
       "      <td>0.998694</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000009</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00032</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>0.000187</td>\n",
       "      <td>0.998881</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000008</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00032</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>0.000188</td>\n",
       "      <td>0.998881</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000053</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00032</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>0.000284</td>\n",
       "      <td>0.998321</td>\n",
       "      <td>5372</td>\n",
       "      <td>0.000009</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>448</td>\n",
       "      <td>0.00032</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       train_loss  train_acc  n_train  val_loss   val_acc  n_val       lr\n",
       "epoch                                                                    \n",
       "1        0.139630   0.270336     5372  0.055159  0.606818    448  0.00500\n",
       "2        0.038713   0.742724     5372  0.016581  0.895455    448  0.00500\n",
       "3        0.018557   0.883396     5372  0.008394  0.945455    448  0.00500\n",
       "4        0.010794   0.923507     5372  0.007102  0.936364    448  0.00500\n",
       "5        0.008575   0.943284     5372  0.007405  0.940909    448  0.00500\n",
       "6        0.008059   0.948321     5372  0.003769  0.968182    448  0.00500\n",
       "7        0.006086   0.960261     5372  0.001893  0.988636    448  0.00500\n",
       "8        0.005367   0.966604     5372  0.002900  0.984091    448  0.00500\n",
       "9        0.003831   0.979104     5372  0.000903  0.990909    448  0.00500\n",
       "10       0.004822   0.972201     5372  0.003475  0.979545    448  0.00500\n",
       "11       0.004310   0.977239     5372  0.001040  0.995455    448  0.00500\n",
       "12       0.001392   0.992164     5372  0.000287  0.997727    448  0.00200\n",
       "13       0.001008   0.993657     5372  0.000182  1.000000    448  0.00200\n",
       "14       0.001024   0.994216     5372  0.001981  0.990909    448  0.00200\n",
       "15       0.001141   0.994590     5372  0.000243  0.997727    448  0.00200\n",
       "16       0.001066   0.994403     5372  0.000198  1.000000    448  0.00200\n",
       "17       0.000646   0.996642     5372  0.000148  1.000000    448  0.00200\n",
       "18       0.000872   0.995709     5372  0.002752  0.979545    448  0.00200\n",
       "19       0.001856   0.989552     5372  0.002104  0.986364    448  0.00200\n",
       "20       0.000949   0.994590     5372  0.000401  0.997727    448  0.00200\n",
       "21       0.001109   0.992724     5372  0.000084  1.000000    448  0.00200\n",
       "22       0.000780   0.996269     5372  0.000117  1.000000    448  0.00200\n",
       "23       0.000548   0.997388     5372  0.000151  0.997727    448  0.00080\n",
       "24       0.000388   0.998134     5372  0.000259  0.997727    448  0.00080\n",
       "25       0.000371   0.997388     5372  0.000033  1.000000    448  0.00080\n",
       "26       0.000491   0.997201     5372  0.000047  1.000000    448  0.00080\n",
       "27       0.000317   0.998507     5372  0.000050  1.000000    448  0.00080\n",
       "28       0.000257   0.998881     5372  0.000327  0.997727    448  0.00080\n",
       "29       0.000304   0.997761     5372  0.000030  1.000000    448  0.00080\n",
       "30       0.000374   0.997761     5372  0.000024  1.000000    448  0.00080\n",
       "31       0.000371   0.997761     5372  0.000085  1.000000    448  0.00080\n",
       "32       0.000452   0.997575     5372  0.000048  1.000000    448  0.00080\n",
       "33       0.000218   0.998694     5372  0.000011  1.000000    448  0.00080\n",
       "34       0.000130   0.999254     5372  0.000010  1.000000    448  0.00032\n",
       "35       0.000176   0.999254     5372  0.000017  1.000000    448  0.00032\n",
       "36       0.000200   0.998694     5372  0.000062  1.000000    448  0.00032\n",
       "37       0.000201   0.998694     5372  0.000009  1.000000    448  0.00032\n",
       "38       0.000187   0.998881     5372  0.000008  1.000000    448  0.00032\n",
       "39       0.000188   0.998881     5372  0.000053  1.000000    448  0.00032\n",
       "40       0.000284   0.998321     5372  0.000009  1.000000    448  0.00032"
      ]
     },
     "execution_count": 89,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "trainer.train(40)\n",
    "trainer.complete_results"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3af51111-f15f-42ab-a11d-2d2b2f314198",
   "metadata": {},
   "source": [
    "Training and inference is quicker than the CNN, but the difference less great from time spent during feature extraction. This model closes the accuracy gap greatly, but doesn't quite reach the accuracy of the VGG16 model with transfer learning applied."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "id": "6f98ffa4-45b1-4670-87e7-803da3590761",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       test_loss  test_acc  n_test       lr\n",
      "epoch                                      \n",
      "40      0.001016  0.996591   896.0  0.00032\n",
      "CPU times: user 56.3 ms, sys: 276 ms, total: 332 ms\n",
      "Wall time: 616 ms\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>test_loss</th>\n",
       "      <th>test_acc</th>\n",
       "      <th>n_test</th>\n",
       "      <th>lr</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>epoch</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>0.001016</td>\n",
       "      <td>0.996591</td>\n",
       "      <td>896.0</td>\n",
       "      <td>0.00032</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       test_loss  test_acc  n_test       lr\n",
       "epoch                                      \n",
       "40      0.001016  0.996591   896.0  0.00032"
      ]
     },
     "execution_count": 90,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "trainer.test()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "id": "152f3e04-fda9-4125-b540-672e87344635",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1400x700 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, [ax1, ax2] = common_plot_setup(style=\"fivethirtyeight\", font=\"DejaVu Sans\", figsize=(14, 7), rows=1, cols=2)\n",
    "trainer.complete_results[[\"train_loss\", \"val_loss\"]].plot(ax=ax1, logy=True)\n",
    "trainer.complete_results[[\"train_acc\", \"val_acc\"]].plot(ax=ax2, logy=False)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9d544459-eeac-4366-8255-4c51a904e500",
   "metadata": {},
   "source": [
    "By examining the outputs when the predicted labels don't match the actual, we can find 3 characters within the dataset to examine in greater detail. The mistaken observations for し and ぞ were very close class among similar shapes, but what was interesting was the misclassification of こ did not even have the correct character in the top 5."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "id": "034c9d51-2409-4125-8fd3-3d575f843804",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Characters missed: 3\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>label</th>\n",
       "      <th>probability</th>\n",
       "      <th>char_id</th>\n",
       "      <th>character</th>\n",
       "      <th>romanization</th>\n",
       "      <th>stroke_count</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <th>label</th>\n",
       "      <th>char_id</th>\n",
       "      <th>character</th>\n",
       "      <th>romanization</th>\n",
       "      <th>indices</th>\n",
       "      <th>stroke_count</th>\n",
       "      <th>strokes_found</th>\n",
       "      <th>pred</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"15\" valign=\"top\">1</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">hir_ko</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">ko</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">こ</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">ko</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">0</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">2</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">2.0</th>\n",
       "      <th>1</th>\n",
       "      <td>hir_e</td>\n",
       "      <td>0.999464</td>\n",
       "      <td>e</td>\n",
       "      <td>え</td>\n",
       "      <td>e</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>hir_u</td>\n",
       "      <td>0.000268</td>\n",
       "      <td>u</td>\n",
       "      <td>う</td>\n",
       "      <td>u</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hir_ra</td>\n",
       "      <td>0.000177</td>\n",
       "      <td>ra</td>\n",
       "      <td>ら</td>\n",
       "      <td>ra</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>hir_to</td>\n",
       "      <td>0.000018</td>\n",
       "      <td>to</td>\n",
       "      <td>と</td>\n",
       "      <td>to</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hir_zo</td>\n",
       "      <td>0.000014</td>\n",
       "      <td>zo</td>\n",
       "      <td>ぞ</td>\n",
       "      <td>zo</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">hir_si</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">si</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">し</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">shi</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">1</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">1</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">1.0</th>\n",
       "      <th>1</th>\n",
       "      <td>hir_n</td>\n",
       "      <td>0.557691</td>\n",
       "      <td>n</td>\n",
       "      <td>ん</td>\n",
       "      <td>n</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>hir_si</td>\n",
       "      <td>0.388516</td>\n",
       "      <td>si</td>\n",
       "      <td>し</td>\n",
       "      <td>shi</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hir_ku</td>\n",
       "      <td>0.034469</td>\n",
       "      <td>ku</td>\n",
       "      <td>く</td>\n",
       "      <td>ku</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>hir_ha</td>\n",
       "      <td>0.016394</td>\n",
       "      <td>ha</td>\n",
       "      <td>は</td>\n",
       "      <td>ha</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hir_ru</td>\n",
       "      <td>0.001736</td>\n",
       "      <td>ru</td>\n",
       "      <td>る</td>\n",
       "      <td>ru</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">hir_zo</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">zo</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">ぞ</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">zo</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">2</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">3</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">3.0</th>\n",
       "      <th>1</th>\n",
       "      <td>hir_de</td>\n",
       "      <td>0.581675</td>\n",
       "      <td>de</td>\n",
       "      <td>で</td>\n",
       "      <td>de</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>hir_zo</td>\n",
       "      <td>0.416614</td>\n",
       "      <td>zo</td>\n",
       "      <td>ぞ</td>\n",
       "      <td>zo</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hir_bi</td>\n",
       "      <td>0.000636</td>\n",
       "      <td>bi</td>\n",
       "      <td>び</td>\n",
       "      <td>bi</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>hir_i</td>\n",
       "      <td>0.000471</td>\n",
       "      <td>i</td>\n",
       "      <td>い</td>\n",
       "      <td>i</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hir_du</td>\n",
       "      <td>0.000161</td>\n",
       "      <td>du</td>\n",
       "      <td>づ</td>\n",
       "      <td>dzu</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                                      label  \\\n",
       "count label  char_id character romanization indices stroke_count strokes_found pred           \n",
       "1     hir_ko ko      こ         ko           0       2            2.0           1      hir_e   \n",
       "                                                                               2      hir_u   \n",
       "                                                                               3     hir_ra   \n",
       "                                                                               4     hir_to   \n",
       "                                                                               5     hir_zo   \n",
       "      hir_si si      し         shi          1       1            1.0           1      hir_n   \n",
       "                                                                               2     hir_si   \n",
       "                                                                               3     hir_ku   \n",
       "                                                                               4     hir_ha   \n",
       "                                                                               5     hir_ru   \n",
       "      hir_zo zo      ぞ         zo           2       3            3.0           1     hir_de   \n",
       "                                                                               2     hir_zo   \n",
       "                                                                               3     hir_bi   \n",
       "                                                                               4      hir_i   \n",
       "                                                                               5     hir_du   \n",
       "\n",
       "                                                                                     probability  \\\n",
       "count label  char_id character romanization indices stroke_count strokes_found pred                \n",
       "1     hir_ko ko      こ         ko           0       2            2.0           1        0.999464   \n",
       "                                                                               2        0.000268   \n",
       "                                                                               3        0.000177   \n",
       "                                                                               4        0.000018   \n",
       "                                                                               5        0.000014   \n",
       "      hir_si si      し         shi          1       1            1.0           1        0.557691   \n",
       "                                                                               2        0.388516   \n",
       "                                                                               3        0.034469   \n",
       "                                                                               4        0.016394   \n",
       "                                                                               5        0.001736   \n",
       "      hir_zo zo      ぞ         zo           2       3            3.0           1        0.581675   \n",
       "                                                                               2        0.416614   \n",
       "                                                                               3        0.000636   \n",
       "                                                                               4        0.000471   \n",
       "                                                                               5        0.000161   \n",
       "\n",
       "                                                                                    char_id  \\\n",
       "count label  char_id character romanization indices stroke_count strokes_found pred           \n",
       "1     hir_ko ko      こ         ko           0       2            2.0           1          e   \n",
       "                                                                               2          u   \n",
       "                                                                               3         ra   \n",
       "                                                                               4         to   \n",
       "                                                                               5         zo   \n",
       "      hir_si si      し         shi          1       1            1.0           1          n   \n",
       "                                                                               2         si   \n",
       "                                                                               3         ku   \n",
       "                                                                               4         ha   \n",
       "                                                                               5         ru   \n",
       "      hir_zo zo      ぞ         zo           2       3            3.0           1         de   \n",
       "                                                                               2         zo   \n",
       "                                                                               3         bi   \n",
       "                                                                               4          i   \n",
       "                                                                               5         du   \n",
       "\n",
       "                                                                                    character  \\\n",
       "count label  char_id character romanization indices stroke_count strokes_found pred             \n",
       "1     hir_ko ko      こ         ko           0       2            2.0           1            え   \n",
       "                                                                               2            う   \n",
       "                                                                               3            ら   \n",
       "                                                                               4            と   \n",
       "                                                                               5            ぞ   \n",
       "      hir_si si      し         shi          1       1            1.0           1            ん   \n",
       "                                                                               2            し   \n",
       "                                                                               3            く   \n",
       "                                                                               4            は   \n",
       "                                                                               5            る   \n",
       "      hir_zo zo      ぞ         zo           2       3            3.0           1            で   \n",
       "                                                                               2            ぞ   \n",
       "                                                                               3            び   \n",
       "                                                                               4            い   \n",
       "                                                                               5            づ   \n",
       "\n",
       "                                                                                    romanization  \\\n",
       "count label  char_id character romanization indices stroke_count strokes_found pred                \n",
       "1     hir_ko ko      こ         ko           0       2            2.0           1               e   \n",
       "                                                                               2               u   \n",
       "                                                                               3              ra   \n",
       "                                                                               4              to   \n",
       "                                                                               5              zo   \n",
       "      hir_si si      し         shi          1       1            1.0           1               n   \n",
       "                                                                               2             shi   \n",
       "                                                                               3              ku   \n",
       "                                                                               4              ha   \n",
       "                                                                               5              ru   \n",
       "      hir_zo zo      ぞ         zo           2       3            3.0           1              de   \n",
       "                                                                               2              zo   \n",
       "                                                                               3              bi   \n",
       "                                                                               4               i   \n",
       "                                                                               5             dzu   \n",
       "\n",
       "                                                                                    stroke_count  \n",
       "count label  char_id character romanization indices stroke_count strokes_found pred               \n",
       "1     hir_ko ko      こ         ko           0       2            2.0           1               2  \n",
       "                                                                               2               2  \n",
       "                                                                               3               2  \n",
       "                                                                               4               2  \n",
       "                                                                               5               3  \n",
       "      hir_si si      し         shi          1       1            1.0           1               1  \n",
       "                                                                               2               1  \n",
       "                                                                               3               1  \n",
       "                                                                               4               3  \n",
       "                                                                               5               1  \n",
       "      hir_zo zo      ぞ         zo           2       3            3.0           1               3  \n",
       "                                                                               2               3  \n",
       "                                                                               3               3  \n",
       "                                                                               4               2  \n",
       "                                                                               5               3  "
      ]
     },
     "execution_count": 92,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "misses = IncorrectCharacters(trainer, conformed_dataset, data_paths.get_classes())\n",
    "print(\"Characters missed:\", len(misses))\n",
    "misses.overview(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2ebd9311-97db-4b68-840e-e0f3ddd14361",
   "metadata": {},
   "source": [
    "Examining it in more detail, it can be seen that the shape and stroke order is technically correct, it is a bit different from the rest of the dataset due to being drawn by a different contributor whose presence was rare in the dataset. The model would have much to gain from an increased diversity of contributers because the current data is made entirely from the contributions of two individuals where one person was responsible for roughly 98% of the observations."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "id": "31854006-5ad5-47b0-8c9c-0fde1584f7a9",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(    label char_id character romanization stroke_count  strokes_found\n",
      "9  hir_ko      ko         こ           ko            2              2,      label  probability char_id character romanization stroke_count\n",
      "3    hir_e     0.999464       e         え            e            2\n",
      "2    hir_u     0.000268       u         う            u            2\n",
      "38  hir_ra     0.000177      ra         ら           ra            2\n",
      "19  hir_to     0.000018      to         と           to            2\n",
      "55  hir_zo     0.000014      zo         ぞ           zo            3)\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0bf2de0591a44fe88e840555f7acbe93",
       "version_major": 2,
       "version_minor": 0
      },
      "image/png": "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",
      "text/html": [
       "\n",
       "            <div style=\"display: inline-block;\">\n",
       "                <div class=\"jupyter-widgets widget-label\" style=\"text-align: center;\">\n",
       "                    Figure\n",
       "                </div>\n",
       "                <img src='data:image/png;base64,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' width=600.0/>\n",
       "            </div>\n",
       "        "
      ],
      "text/plain": [
       "Canvas(toolbar=Toolbar(toolitems=[('Home', 'Reset original view', 'home', 'home'), ('Back', 'Back to previous …"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "    label char_id character romanization stroke_count  strokes_found\n",
      "9  hir_ko      ko         こ           ko            2              2\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'../data/choubenkyo_data/hir_ko_2_kay_0000.json'"
      ]
     },
     "execution_count": 95,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%matplotlib widget\n",
    "i = 0\n",
    "print(misses.compare(i, 5))\n",
    "misses.examine(i, force_stop_previous=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1d81ef4b-2bd2-4f68-87b2-0289c429734c",
   "metadata": {},
   "source": [
    "## Conclusion"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "840971dc-0a37-426d-8e0b-840b9e5f9ae6",
   "metadata": {},
   "source": [
    "The refinements to the process have greatly improved accuracy in comparison to the earlier prototype.\n",
    "\n",
    "Based on the results in this notebook and the advantages and disadvantages presented, there is strong incentive to further develop the idea of model 4 as the strategy in the event of a continued project and expanded scope. While the VGG16 transfer learned CNN was impressive and outclassed the other models, the ability to fully utilize the sequential data would likely outpace it under the assumption of an expanded dataset and classes and outperform its implementation.\n",
    "\n",
    "The next step for a more robust model would likely be best served in exploring additional refinements to the training process. Most notably diversifying away from padding all observations to a fixed amount of strokes, but instead dynamically training complete unpadded characters augmented by clipped and padded variations that maintain a reasonable balance of classes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b03e8368-0e07-4d24-be98-c33519b7a6eb",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "choubenkyo",
   "language": "python",
   "name": "choubenkyo"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}