Keras + Tensorflow MNIST.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Keras + Tensorflow implementation of handwritten digits recognition"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Data from Kaggle Digit Recognition competition  \n",
    "https://www.kaggle.com/c/digit-recognizer/data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Activation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/Users/zubo/Dropbox/projects/kaggle/digit-recognizer\r\n"
     ]
    }
   ],
   "source": [
    "!pwd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "train_file = 'data/train.csv'\n",
    "test_file = 'data/test.csv'\n",
    "\n",
    "train_df = pd.read_csv(train_file)\n",
    "test_df = pd.read_csv(test_file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>label</th>\n",
       "      <th>pixel0</th>\n",
       "      <th>pixel1</th>\n",
       "      <th>pixel2</th>\n",
       "      <th>pixel3</th>\n",
       "      <th>pixel4</th>\n",
       "      <th>pixel5</th>\n",
       "      <th>pixel6</th>\n",
       "      <th>pixel7</th>\n",
       "      <th>pixel8</th>\n",
       "      <th>...</th>\n",
       "      <th>pixel774</th>\n",
       "      <th>pixel775</th>\n",
       "      <th>pixel776</th>\n",
       "      <th>pixel777</th>\n",
       "      <th>pixel778</th>\n",
       "      <th>pixel779</th>\n",
       "      <th>pixel780</th>\n",
       "      <th>pixel781</th>\n",
       "      <th>pixel782</th>\n",
       "      <th>pixel783</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 785 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   label  pixel0  pixel1  pixel2  pixel3  pixel4  pixel5  pixel6  pixel7  \\\n",
       "0      1       0       0       0       0       0       0       0       0   \n",
       "1      0       0       0       0       0       0       0       0       0   \n",
       "2      1       0       0       0       0       0       0       0       0   \n",
       "3      4       0       0       0       0       0       0       0       0   \n",
       "4      0       0       0       0       0       0       0       0       0   \n",
       "\n",
       "   pixel8    ...     pixel774  pixel775  pixel776  pixel777  pixel778  \\\n",
       "0       0    ...            0         0         0         0         0   \n",
       "1       0    ...            0         0         0         0         0   \n",
       "2       0    ...            0         0         0         0         0   \n",
       "3       0    ...            0         0         0         0         0   \n",
       "4       0    ...            0         0         0         0         0   \n",
       "\n",
       "   pixel779  pixel780  pixel781  pixel782  pixel783  \n",
       "0         0         0         0         0         0  \n",
       "1         0         0         0         0         0  \n",
       "2         0         0         0         0         0  \n",
       "3         0         0         0         0         0  \n",
       "4         0         0         0         0         0  \n",
       "\n",
       "[5 rows x 785 columns]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df.head() # labels and pixels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>pixel0</th>\n",
       "      <th>pixel1</th>\n",
       "      <th>pixel2</th>\n",
       "      <th>pixel3</th>\n",
       "      <th>pixel4</th>\n",
       "      <th>pixel5</th>\n",
       "      <th>pixel6</th>\n",
       "      <th>pixel7</th>\n",
       "      <th>pixel8</th>\n",
       "      <th>pixel9</th>\n",
       "      <th>...</th>\n",
       "      <th>pixel774</th>\n",
       "      <th>pixel775</th>\n",
       "      <th>pixel776</th>\n",
       "      <th>pixel777</th>\n",
       "      <th>pixel778</th>\n",
       "      <th>pixel779</th>\n",
       "      <th>pixel780</th>\n",
       "      <th>pixel781</th>\n",
       "      <th>pixel782</th>\n",
       "      <th>pixel783</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 784 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   pixel0  pixel1  pixel2  pixel3  pixel4  pixel5  pixel6  pixel7  pixel8  \\\n",
       "0       0       0       0       0       0       0       0       0       0   \n",
       "1       0       0       0       0       0       0       0       0       0   \n",
       "2       0       0       0       0       0       0       0       0       0   \n",
       "3       0       0       0       0       0       0       0       0       0   \n",
       "4       0       0       0       0       0       0       0       0       0   \n",
       "\n",
       "   pixel9    ...     pixel774  pixel775  pixel776  pixel777  pixel778  \\\n",
       "0       0    ...            0         0         0         0         0   \n",
       "1       0    ...            0         0         0         0         0   \n",
       "2       0    ...            0         0         0         0         0   \n",
       "3       0    ...            0         0         0         0         0   \n",
       "4       0    ...            0         0         0         0         0   \n",
       "\n",
       "   pixel779  pixel780  pixel781  pixel782  pixel783  \n",
       "0         0         0         0         0         0  \n",
       "1         0         0         0         0         0  \n",
       "2         0         0         0         0         0  \n",
       "3         0         0         0         0         0  \n",
       "4         0         0         0         0         0  \n",
       "\n",
       "[5 rows x 784 columns]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_df.head() # no labels here"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So each row of the dataset is a sequence of ink intensity for the 784 pixels that represents square image with the size 28x28."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "train_labels = train_df.label\n",
    "train_images = train_df.iloc[:,1:]\n",
    "test_images = test_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Show some digits from the input dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAsMAAAFUCAYAAADBBzEkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1155dd9e8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(12,6))\n",
    "for i in range(0,9):\n",
    "    plt.subplot(250 + (i+1))\n",
    "    img = train_images.ix[i,:].values.reshape(28, 28)\n",
    "    plt.imshow(img, cmap='Greys')\n",
    "    plt.title(train_labels[i])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preprocess data for NN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "train_images = (train_images/train_images.max()).fillna(0) # normalize values\n",
    "test_images = (test_images/test_images.max()).fillna(0) # normalize values\n",
    "train_labels = pd.get_dummies(train_labels) # one-hot encoding of the label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   0  1  2  3  4  5  6  7  8  9\n",
       "0  0  1  0  0  0  0  0  0  0  0\n",
       "1  1  0  0  0  0  0  0  0  0  0\n",
       "2  0  1  0  0  0  0  0  0  0  0\n",
       "3  0  0  0  0  1  0  0  0  0  0\n",
       "4  1  0  0  0  0  0  0  0  0  0"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_labels.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Build simple 1-hidden layer dense NN\n",
    "Some more information about Keras' Sequential models  \n",
    "https://keras.io/getting-started/sequential-model-guide/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model = Sequential([\n",
    "    Dense(32, input_dim=784),\n",
    "    Activation('relu'),\n",
    "    Dense(10),\n",
    "    Activation('softmax')\n",
    "])\n",
    "\n",
    "# Multi-class classification problem\n",
    "model.compile(optimizer='rmsprop',\n",
    "              loss='categorical_crossentropy',\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "____________________________________________________________________________________________________\n",
      "Layer (type)                     Output Shape          Param #     Connected to                     \n",
      "====================================================================================================\n",
      "dense_1 (Dense)                  (None, 32)            25120       dense_input_1[0][0]              \n",
      "____________________________________________________________________________________________________\n",
      "activation_1 (Activation)        (None, 32)            0           dense_1[0][0]                    \n",
      "____________________________________________________________________________________________________\n",
      "dense_2 (Dense)                  (None, 10)            330         activation_1[0][0]               \n",
      "____________________________________________________________________________________________________\n",
      "activation_2 (Activation)        (None, 10)            0           dense_2[0][0]                    \n",
      "====================================================================================================\n",
      "Total params: 25450\n",
      "____________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# import pydot\n",
    "# from keras.utils.visualize_util import plot\n",
    "# plot(model, to_file='model.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(42000, 784)\n",
      "(42000, 10)\n"
     ]
    }
   ],
   "source": [
    "print(train_images.values.shape)\n",
    "print(train_labels.values.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's fit the weights of our NN. To estimate overfitting, let's use validation dataset which we set up as 5% of the training dataset - validation_split = 0.05."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 39900 samples, validate on 2100 samples\n",
      "Epoch 1/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.4525 - acc: 0.8794 - val_loss: 0.2920 - val_acc: 0.9214\n",
      "Epoch 2/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.2330 - acc: 0.9328 - val_loss: 0.2419 - val_acc: 0.9343\n",
      "Epoch 3/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.1889 - acc: 0.9452 - val_loss: 0.2206 - val_acc: 0.9400\n",
      "Epoch 4/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.1626 - acc: 0.9526 - val_loss: 0.1972 - val_acc: 0.9457\n",
      "Epoch 5/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.1437 - acc: 0.9573 - val_loss: 0.1882 - val_acc: 0.9495\n",
      "Epoch 6/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.1297 - acc: 0.9625 - val_loss: 0.1775 - val_acc: 0.9529\n",
      "Epoch 7/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.1198 - acc: 0.9650 - val_loss: 0.1717 - val_acc: 0.9567\n",
      "Epoch 8/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.1103 - acc: 0.9689 - val_loss: 0.1700 - val_acc: 0.9533\n",
      "Epoch 9/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.1032 - acc: 0.9695 - val_loss: 0.1628 - val_acc: 0.9586\n",
      "Epoch 10/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.0970 - acc: 0.9714 - val_loss: 0.1732 - val_acc: 0.9533\n",
      "Epoch 11/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.0916 - acc: 0.9731 - val_loss: 0.1709 - val_acc: 0.9581\n",
      "Epoch 12/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.0869 - acc: 0.9749 - val_loss: 0.1746 - val_acc: 0.9548\n",
      "Epoch 13/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.0824 - acc: 0.9765 - val_loss: 0.1611 - val_acc: 0.9581\n",
      "Epoch 14/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.0793 - acc: 0.9770 - val_loss: 0.1693 - val_acc: 0.9590\n",
      "Epoch 15/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.0753 - acc: 0.9781 - val_loss: 0.1699 - val_acc: 0.9571\n",
      "Epoch 16/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.0724 - acc: 0.9794 - val_loss: 0.1640 - val_acc: 0.9586\n",
      "Epoch 17/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.0684 - acc: 0.9807 - val_loss: 0.1826 - val_acc: 0.9586\n",
      "Epoch 18/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.0661 - acc: 0.9813 - val_loss: 0.1702 - val_acc: 0.9571\n",
      "Epoch 19/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.0633 - acc: 0.9820 - val_loss: 0.1716 - val_acc: 0.9614\n",
      "Epoch 20/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.0614 - acc: 0.9819 - val_loss: 0.1742 - val_acc: 0.9567\n",
      "Epoch 21/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.0587 - acc: 0.9833 - val_loss: 0.1780 - val_acc: 0.9595\n",
      "Epoch 22/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.0568 - acc: 0.9845 - val_loss: 0.1842 - val_acc: 0.9600\n",
      "Epoch 23/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.0551 - acc: 0.9846 - val_loss: 0.1708 - val_acc: 0.9581\n",
      "Epoch 24/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.0530 - acc: 0.9856 - val_loss: 0.1734 - val_acc: 0.9576\n",
      "Epoch 25/25\n",
      "39900/39900 [==============================] - 2s - loss: 0.0510 - acc: 0.9858 - val_loss: 0.1784 - val_acc: 0.9629\n"
     ]
    }
   ],
   "source": [
    "history=model.fit(train_images.values, train_labels.values, validation_split = 0.05, \n",
    "            nb_epoch=25, batch_size=64)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "hist_df = pd.DataFrame(history.history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>acc</th>\n",
       "      <th>loss</th>\n",
       "      <th>val_acc</th>\n",
       "      <th>val_loss</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.879398</td>\n",
       "      <td>0.452548</td>\n",
       "      <td>0.921429</td>\n",
       "      <td>0.291974</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.932782</td>\n",
       "      <td>0.232999</td>\n",
       "      <td>0.934286</td>\n",
       "      <td>0.241881</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.945213</td>\n",
       "      <td>0.188870</td>\n",
       "      <td>0.940000</td>\n",
       "      <td>0.220622</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.952556</td>\n",
       "      <td>0.162578</td>\n",
       "      <td>0.945714</td>\n",
       "      <td>0.197196</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.957268</td>\n",
       "      <td>0.143722</td>\n",
       "      <td>0.949524</td>\n",
       "      <td>0.188219</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        acc      loss   val_acc  val_loss\n",
       "0  0.879398  0.452548  0.921429  0.291974\n",
       "1  0.932782  0.232999  0.934286  0.241881\n",
       "2  0.945213  0.188870  0.940000  0.220622\n",
       "3  0.952556  0.162578  0.945714  0.197196\n",
       "4  0.957268  0.143722  0.949524  0.188219"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hist_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x11552de80>"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA0oAAAGECAYAAAAIp9AAAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1154440b8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(14,6))\n",
    "plt.style.use('bmh')\n",
    "params_dict = dict(linestyle='solid', linewidth=0.25, marker='o', markersize=6)\n",
    "\n",
    "plt.subplot(121)\n",
    "plt.plot(hist_df.loss, label='Training loss', **params_dict)\n",
    "plt.plot(hist_df.val_loss, label='Validation loss', **params_dict)\n",
    "plt.title('Loss for ' + str(len(history.epoch)) + ' epochs')\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.subplot(122)\n",
    "plt.plot(hist_df.acc, label='Training accuracy', **params_dict)\n",
    "plt.plot(hist_df.val_acc, label='Validation accuracy', **params_dict)\n",
    "plt.title('Accuracy for ' + str(len(history.epoch)) + ' epochs')\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Some conclusions here: \n",
    "1. At some number of the epochs (8-10 in this case) the validation loss stops decreasing, i.e. the network begin to overfit. The training loss still decreases.\n",
    "2. The same is true for the training/validation accuracy - validation accuracy stops increasing after some number of epochs.\n",
    "3. Since validation data was not used for training, measuring accuracy on the the validation dataset gives us an estimation of the ability of the model to generalize."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Basically we tuned the only one parameter - number of epochs using validation.\n",
    "Now I will use epochs=10."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "42000/42000 [==============================] - 3s - loss: 0.4534 - acc: 0.8757     \n",
      "Epoch 2/10\n",
      "42000/42000 [==============================] - 2s - loss: 0.2473 - acc: 0.9294     \n",
      "Epoch 3/10\n",
      "42000/42000 [==============================] - 3s - loss: 0.2030 - acc: 0.9400     \n",
      "Epoch 4/10\n",
      "42000/42000 [==============================] - 3s - loss: 0.1744 - acc: 0.9490     \n",
      "Epoch 5/10\n",
      "42000/42000 [==============================] - 3s - loss: 0.1537 - acc: 0.9545     \n",
      "Epoch 6/10\n",
      "42000/42000 [==============================] - 3s - loss: 0.1385 - acc: 0.9594     \n",
      "Epoch 7/10\n",
      "42000/42000 [==============================] - 3s - loss: 0.1266 - acc: 0.9634     \n",
      "Epoch 8/10\n",
      "42000/42000 [==============================] - 3s - loss: 0.1168 - acc: 0.9651     \n",
      "Epoch 9/10\n",
      "42000/42000 [==============================] - 3s - loss: 0.1076 - acc: 0.9679     \n",
      "Epoch 10/10\n",
      "42000/42000 [==============================] - 3s - loss: 0.1008 - acc: 0.9700     \n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x105b63588>"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Predict on the whole dataset now, 10 epochs\n",
    "model = Sequential([\n",
    "    Dense(32, input_dim=784),\n",
    "    Activation('relu'),\n",
    "    Dense(10),\n",
    "    Activation('softmax')\n",
    "])\n",
    "\n",
    "# Multi-class classification problem\n",
    "model.compile(optimizer='rmsprop',\n",
    "              loss='categorical_crossentropy',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "model.fit(train_images.values, train_labels.values, nb_epoch=10, batch_size=64)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Predict test labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "27904/28000 [============================>.] - ETA: 0s"
     ]
    }
   ],
   "source": [
    "pred_classes = model.predict_classes(test_images.values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([2, 0, 9, ..., 3, 9, 2])"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_classes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "pred = pd.DataFrame({'ImageId': range(1, len(pred_classes)+1), 'Label': pred_classes})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>ImageId</th>\n",
       "      <th>Label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   ImageId  Label\n",
       "0        1      2\n",
       "1        2      0\n",
       "2        3      9\n",
       "3        4      9\n",
       "4        5      3"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>ImageId</th>\n",
       "      <th>Label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>27995</th>\n",
       "      <td>27996</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27996</th>\n",
       "      <td>27997</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27997</th>\n",
       "      <td>27998</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27998</th>\n",
       "      <td>27999</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27999</th>\n",
       "      <td>28000</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       ImageId  Label\n",
       "27995    27996      9\n",
       "27996    27997      7\n",
       "27997    27998      3\n",
       "27998    27999      9\n",
       "27999    28000      2"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred.tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "pred.to_csv('data/output/subm06.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [conda env:tensorflow]",
   "language": "python",
   "name": "conda-env-tensorflow-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}