RL_DP_Demo.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "RL_DP_Demo.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/AmanPriyanshu/DP-HyperparamTuning/blob/main/RL_DP_Demo.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "p5E4oLWCcJSd"
      },
      "source": [
        "# Demo for RL-DP-Project:\n",
        "\n",
        "## SET-UP"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xgxkz13IX_W9",
        "outputId": "72486692-34ce-4d24-9fbf-c1e18fb45637"
      },
      "source": [
        "!nvidia-smi"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "3g0ts-lKYMXG",
        "outputId": "bdfa6adb-aed2-4597-903e-48e3aa1ddcb9"
      },
      "source": [
        "!git clone https://github.com/AmanPriyanshu/DP-HyperparamTuning.git ./RL_DP_Project"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Cloning into './RL_DP_Project'...\n",
            "remote: Enumerating objects: 98, done.\u001b[K\n",
            "remote: Counting objects: 100% (98/98), done.\u001b[K\n",
            "remote: Compressing objects: 100% (80/80), done.\u001b[K\n",
            "remote: Total 98 (delta 36), reused 48 (delta 12), pack-reused 0\u001b[K\n",
            "Unpacking objects: 100% (98/98), done.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jbVvSy_6cbLL"
      },
      "source": [
        "## Importing Everything:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "sOLdsmkkchpL",
        "outputId": "95fa1b67-775a-4dc0-b7e4-ccf87ec9e264"
      },
      "source": [
        "!pip install opacus"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting opacus\n",
            "  Downloading opacus-0.14.0-py3-none-any.whl (114 kB)\n",
            "\u001b[?25l\r\u001b[K     |██▉                             | 10 kB 10.5 MB/s eta 0:00:01\r\u001b[K     |█████▊                          | 20 kB 14.5 MB/s eta 0:00:01\r\u001b[K     |████████▋                       | 30 kB 10.4 MB/s eta 0:00:01\r\u001b[K     |███████████▌                    | 40 kB 9.3 MB/s eta 0:00:01\r\u001b[K     |██████████████▍                 | 51 kB 5.6 MB/s eta 0:00:01\r\u001b[K     |█████████████████▏              | 61 kB 5.3 MB/s eta 0:00:01\r\u001b[K     |████████████████████            | 71 kB 4.9 MB/s eta 0:00:01\r\u001b[K     |███████████████████████         | 81 kB 5.5 MB/s eta 0:00:01\r\u001b[K     |█████████████████████████▉      | 92 kB 5.5 MB/s eta 0:00:01\r\u001b[K     |████████████████████████████▊   | 102 kB 5.2 MB/s eta 0:00:01\r\u001b[K     |███████████████████████████████▌| 112 kB 5.2 MB/s eta 0:00:01\r\u001b[K     |████████████████████████████████| 114 kB 5.2 MB/s \n",
            "\u001b[?25hRequirement already satisfied: numpy>=1.15 in /usr/local/lib/python3.7/dist-packages (from opacus) (1.19.5)\n",
            "Requirement already satisfied: scipy>=1.2 in /usr/local/lib/python3.7/dist-packages (from opacus) (1.4.1)\n",
            "Requirement already satisfied: torch>=1.3 in /usr/local/lib/python3.7/dist-packages (from opacus) (1.9.0+cu102)\n",
            "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from torch>=1.3->opacus) (3.7.4.3)\n",
            "Installing collected packages: opacus\n",
            "Successfully installed opacus-0.14.0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "G0Oe4TMyceQz"
      },
      "source": [
        "from opacus import PrivacyEngine\n",
        "from tqdm.notebook import tqdm\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "import torch\n",
        "import random\n",
        "import os\n",
        "from sklearn.datasets import make_classification\n",
        "import warnings\n",
        "warnings.filterwarnings(\"ignore\")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1L_OUEmxcyMU"
      },
      "source": [
        "## Reading the Data:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_2g9fg1ec1Aj"
      },
      "source": [
        "class ClassificationDataset(torch.utils.data.Dataset):\n",
        "\tdef __init__(self, x, y):\n",
        "\t\tself.x = x\n",
        "\t\tself.y = y\n",
        "\n",
        "\tdef __len__(self):\n",
        "\t\treturn len(self.x)\n",
        "\n",
        "\tdef __getitem__(self, idx):\n",
        "\t\treturn torch.from_numpy(self.x[idx].astype(np.float32)), torch.from_numpy(np.array([self.y[idx]]).astype(np.float32))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yn1xnKAzc-iK"
      },
      "source": [
        "def load_sklearn(val_split=0.2):\n",
        "\tx, y = make_classification(n_samples=4000, n_features=8, n_informative=2, n_redundant=2, n_classes=2, n_clusters_per_class=2, flip_y=0.15, class_sep=1.5, hypercube=True, shift=0.0, shuffle=True, random_state=0)\n",
        "\ttrain_x, train_y, test_x, test_y = x[:int((1-val_split)*len(x))], y[:int((1-val_split)*len(x))], x[int((1-val_split)*len(x)):], y[int((1-val_split)*len(x)):]\n",
        "\treturn train_x, train_y, test_x, test_y"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AiwRAHtldDiK"
      },
      "source": [
        "def load_dataset():\n",
        "\ttrain_x, train_y, test_x, test_y = load_sklearn()\n",
        "\ttrain_dataset = ClassificationDataset(train_x, train_y)\n",
        "\ttest_dataset = ClassificationDataset(test_x, test_y)\n",
        "\treturn train_dataset, test_dataset"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "eAy7OJJ1dYmr",
        "outputId": "c229c357-97c1-468c-d5d3-3a378e272d50"
      },
      "source": [
        "train_dataset, test_dataset = load_dataset()\n",
        "print(\"Training:\", type(train_dataset), \"Size:\", len(train_dataset))\n",
        "print(\"Testing:\", type(test_dataset), \"Size:\", len(test_dataset))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Training: <class '__main__.ClassificationDataset'> Size: 3200\n",
            "Testing: <class '__main__.ClassificationDataset'> Size: 800\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2pejgswmcRl6"
      },
      "source": [
        "## Creating a Torch-Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "N5F6xHuSYjqI"
      },
      "source": [
        "def get_model():\n",
        "\tmodel = torch.nn.Sequential(\n",
        "\t\t\ttorch.nn.Linear(8, 4),\n",
        "\t\t\ttorch.nn.ReLU(),\n",
        "\t\t\ttorch.nn.Linear(4, 1),\n",
        "\t\t\ttorch.nn.Sigmoid(),\n",
        "\t\t)\n",
        "\treturn model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "lBCtQwrddkyj",
        "outputId": "64ca1bc8-5624-4190-e36a-c6f37e91eff7"
      },
      "source": [
        "print(get_model())"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Sequential(\n",
            "  (0): Linear(in_features=8, out_features=4, bias=True)\n",
            "  (1): ReLU()\n",
            "  (2): Linear(in_features=4, out_features=1, bias=True)\n",
            "  (3): Sigmoid()\n",
            ")\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "t_2J3U7zd9Ly"
      },
      "source": [
        "## Loading our Algorithms:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hBvIABoKeCOq"
      },
      "source": [
        "from RL_DP_Project.experiment.train_single_model import Experiment\n",
        "from RL_DP_Project.algorithms.bayesian_optimization import Bayesian\n",
        "from RL_DP_Project.algorithms.grid_search_algorithm import GridSearch\n",
        "from RL_DP_Project.algorithms.evolutionary_optimization import EvolutionaryOptimization\n",
        "from RL_DP_Project.algorithms.reinforcement_learning_optimization import RLOptimization"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Mngh93AkdtnB"
      },
      "source": [
        "## Running A DP-Experiment"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Mc4-i3IPdlZV"
      },
      "source": [
        "def run_sample():\n",
        "\tcriterion = torch.nn.BCELoss()\n",
        "\ttrain_dataset, test_dataset = load_dataset()\n",
        "\te = Experiment(get_model, criterion, train_dataset, test_dataset)\n",
        "\tresults = e.run_experiment(1, 0.001)\n",
        "\tprint()\n",
        "\tprint(\"RESULTS:\")\n",
        "\t_ = [print(key+\":\", round(item, 4)) for key, item in results.items()]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "3pmKVQT-ePnI",
        "outputId": "2831bb57-f749-438c-8415-66796532a826"
      },
      "source": [
        "run_sample()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "{'type': 'training', 'epoch': 1, 'loss': 0.6896, 'acc': 0.5613}: 100%|██████████| 400/400 [00:04<00:00, 85.53it/s]\n",
            "{'type': 'testing', 'epoch': 1, 'loss': 0.6919, 'acc': 0.5563}: 100%|██████████| 100/100 [00:00<00:00, 222.54it/s]\n",
            "{'type': 'training', 'epoch': 2, 'loss': 0.6751, 'acc': 0.5875}: 100%|██████████| 400/400 [00:04<00:00, 84.02it/s]\n",
            "{'type': 'testing', 'epoch': 2, 'loss': 0.6777, 'acc': 0.58}: 100%|██████████| 100/100 [00:00<00:00, 152.66it/s]\n",
            "{'type': 'training', 'epoch': 3, 'loss': 0.6601, 'acc': 0.6134}: 100%|██████████| 400/400 [00:04<00:00, 91.88it/s] \n",
            "{'type': 'testing', 'epoch': 3, 'loss': 0.662, 'acc': 0.595}: 100%|██████████| 100/100 [00:00<00:00, 209.07it/s]\n",
            "{'type': 'training', 'epoch': 4, 'loss': 0.6432, 'acc': 0.6338}: 100%|██████████| 400/400 [00:04<00:00, 81.62it/s]\n",
            "{'type': 'testing', 'epoch': 4, 'loss': 0.6439, 'acc': 0.6112}: 100%|██████████| 100/100 [00:00<00:00, 161.76it/s]\n",
            "{'type': 'training', 'epoch': 5, 'loss': 0.6242, 'acc': 0.6606}: 100%|██████████| 400/400 [00:04<00:00, 89.51it/s]\n",
            "{'type': 'testing', 'epoch': 5, 'loss': 0.6238, 'acc': 0.6438}: 100%|██████████| 100/100 [00:00<00:00, 157.54it/s]\n",
            "{'type': 'training', 'epoch': 6, 'loss': 0.6037, 'acc': 0.6887}: 100%|██████████| 400/400 [00:04<00:00, 85.68it/s]\n",
            "{'type': 'testing', 'epoch': 6, 'loss': 0.6022, 'acc': 0.685}: 100%|██████████| 100/100 [00:00<00:00, 179.08it/s]\n",
            "{'type': 'training', 'epoch': 7, 'loss': 0.581, 'acc': 0.7278}: 100%|██████████| 400/400 [00:04<00:00, 89.14it/s]\n",
            "{'type': 'testing', 'epoch': 7, 'loss': 0.5776, 'acc': 0.7412}: 100%|██████████| 100/100 [00:00<00:00, 195.57it/s]\n",
            "{'type': 'training', 'epoch': 8, 'loss': 0.5551, 'acc': 0.7731}: 100%|██████████| 400/400 [00:05<00:00, 77.13it/s]\n",
            "{'type': 'testing', 'epoch': 8, 'loss': 0.5504, 'acc': 0.7887}: 100%|██████████| 100/100 [00:00<00:00, 174.79it/s]\n",
            "{'type': 'training', 'epoch': 9, 'loss': 0.529, 'acc': 0.8119}: 100%|██████████| 400/400 [00:04<00:00, 87.71it/s]\n",
            "{'type': 'testing', 'epoch': 9, 'loss': 0.523, 'acc': 0.8263}: 100%|██████████| 100/100 [00:00<00:00, 195.90it/s]\n",
            "{'type': 'training', 'epoch': 10, 'loss': 0.5031, 'acc': 0.8366}: 100%|██████████| 400/400 [00:04<00:00, 80.19it/s]\n",
            "{'type': 'testing', 'epoch': 10, 'loss': 0.4962, 'acc': 0.8363}: 100%|██████████| 100/100 [00:00<00:00, 153.33it/s]\n",
            "{'type': 'training', 'epoch': 11, 'loss': 0.4781, 'acc': 0.8538}: 100%|██████████| 400/400 [00:04<00:00, 83.24it/s]\n",
            "{'type': 'testing', 'epoch': 11, 'loss': 0.4695, 'acc': 0.8538}: 100%|██████████| 100/100 [00:00<00:00, 156.07it/s]\n",
            "{'type': 'training', 'epoch': 12, 'loss': 0.453, 'acc': 0.8634}: 100%|██████████| 400/400 [00:04<00:00, 83.57it/s] \n",
            "{'type': 'testing', 'epoch': 12, 'loss': 0.4433, 'acc': 0.8688}: 100%|██████████| 100/100 [00:00<00:00, 197.30it/s]\n",
            "{'type': 'training', 'epoch': 13, 'loss': 0.4298, 'acc': 0.8731}: 100%|██████████| 400/400 [00:04<00:00, 80.97it/s]\n",
            "{'type': 'testing', 'epoch': 13, 'loss': 0.4203, 'acc': 0.8725}: 100%|██████████| 100/100 [00:00<00:00, 152.44it/s]\n",
            "{'type': 'training', 'epoch': 14, 'loss': 0.4104, 'acc': 0.8772}: 100%|██████████| 400/400 [00:04<00:00, 87.35it/s] \n",
            "{'type': 'testing', 'epoch': 14, 'loss': 0.4012, 'acc': 0.8775}: 100%|██████████| 100/100 [00:00<00:00, 216.13it/s]\n",
            "{'type': 'training', 'epoch': 15, 'loss': 0.3941, 'acc': 0.8816}: 100%|██████████| 400/400 [00:04<00:00, 82.54it/s]\n",
            "{'type': 'testing', 'epoch': 15, 'loss': 0.3854, 'acc': 0.8862}: 100%|██████████| 100/100 [00:00<00:00, 195.52it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "RESULTS:\n",
            "eps: 1.2102\n",
            "train_loss: 0.3941\n",
            "val_loss: 0.3854\n",
            "train_acc: 0.8816\n",
            "val_acc: 0.8862\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0LvTqjrIfbh5"
      },
      "source": [
        "## Creating our Reward Function:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "raP_BobKeRNe"
      },
      "source": [
        "def calculate_reward(eps, train_loss, val_loss, alpha_u=0.5, alpha_p=0.5):\n",
        "\treturn alpha_p*np.exp(-(eps)) + alpha_u*np.exp(-(val_loss))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "F1LuhQZrgEWB"
      },
      "source": [
        "## Creating Functions to Run Optimizers:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BeM06-21gI50"
      },
      "source": [
        "def run_grid_search():\n",
        "\tcriterion = torch.nn.BCELoss()\n",
        "\ttrain_dataset, test_dataset = load_dataset()\n",
        "\te = Experiment(get_model, criterion, train_dataset, test_dataset)\n",
        "\tgs = GridSearch(e.run_experiment, calculate_reward, 10, search_space_nm=[2, 5], search_space_lr=[0.001, 0.05])\n",
        "\tprogress = gs.run()\n",
        "\treturn progress[::-1]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KJRL65p-gBHx"
      },
      "source": [
        "def run_bayesian():\n",
        "\tcriterion = torch.nn.BCELoss()\n",
        "\ttrain_dataset, test_dataset = load_dataset()\n",
        "\te = Experiment(get_model, criterion, train_dataset, test_dataset)\n",
        "\tb = Bayesian(e.run_experiment, calculate_reward, 100, search_space_nm=[2, 5], search_space_lr=[0.001, 0.05])\n",
        "\tprogress = b.run()\n",
        "\treturn progress"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MV1sv3QngLnW"
      },
      "source": [
        "def run_evolutionary_optimization():\n",
        "\tcriterion = torch.nn.BCELoss()\n",
        "\ttrain_dataset, test_dataset = load_dataset()\n",
        "\te = Experiment(get_model, criterion, train_dataset, test_dataset)\n",
        "\teo = EvolutionaryOptimization(e.run_experiment, calculate_reward, 10, search_space_nm=[2, 5], search_space_lr=[0.001, 0.05])\n",
        "\tprogress = eo.run()\n",
        "\treturn progress"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UB33mk39gN7O"
      },
      "source": [
        "def run_reinforcement_learning_optimization():\n",
        "\tcriterion = torch.nn.BCELoss()\n",
        "\ttrain_dataset, test_dataset = load_dataset()\n",
        "\te = Experiment(get_model, criterion, train_dataset, test_dataset)\n",
        "\trl = RLOptimization(e.run_experiment, calculate_reward, 10, search_space_nm=[2, 5], search_space_lr=[0.001, 0.05])\n",
        "\tprogress = rl.run()\n",
        "\treturn progress"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EIJVRJllgW-p"
      },
      "source": [
        "## Running Each Optimizer:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5kXYa9vygQ9u",
        "outputId": "09455a60-1f4c-4fb5-f3bc-2ac48d0f29fc"
      },
      "source": [
        "gs_progress = run_grid_search()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "100%|██████████| 100/100 [21:42<00:00, 13.02s/it]\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WZxGDKoKgeR_",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "16a69a19-769b-4f94-ec16-85df1c24fa3b"
      },
      "source": [
        "bo_progress = run_bayesian()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "100%|██████████| 100/100 [21:03<00:00, 12.63s/it, best loss: 0.2147609112295663]\n",
            "{'lr': 0.0017940862990946287, 'nm': 4.800096455988798}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZQ1ajodLhR4z",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "fde1e79a-662d-464d-e43d-e6b33be5d9d6"
      },
      "source": [
        "eo_progress = run_evolutionary_optimization()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 0, 'lr': 0.0206, 'nm': 2.3, 'eps': 0.3326, 'val_loss': 0.4388, 'reward': 0.6809}: 100%|██████████| 10/10 [02:08<00:00, 12.89s/it]\n",
            "  0%|          | 0/9 [00:00<?, ?it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 0, 'reward_mean': 0.6934687832203716, 'reward_max': 0.7575398511773139}\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 1, 'lr': 0.0206, 'nm': 4.1, 'eps': 0.176, 'val_loss': 0.4125, 'reward': 0.7503}: 100%|██████████| 9/9 [01:55<00:00, 12.86s/it]\n",
            "  0%|          | 0/9 [00:00<?, ?it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 1, 'reward_mean': 0.7445561607140635, 'reward_max': 0.7711721586642581}\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 2, 'lr': 0.0353, 'nm': 2.3, 'eps': 0.3326, 'val_loss': 0.4431, 'reward': 0.6796}: 100%|██████████| 9/9 [01:56<00:00, 12.94s/it]\n",
            "  0%|          | 0/9 [00:00<?, ?it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 2, 'reward_mean': 0.7084664903838429, 'reward_max': 0.7687385396624913}\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 3, 'lr': 0.0402, 'nm': 3.8, 'eps': 0.1885, 'val_loss': 0.6649, 'reward': 0.6713}: 100%|██████████| 9/9 [01:58<00:00, 13.16s/it]\n",
            "  0%|          | 0/9 [00:00<?, ?it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 3, 'reward_mean': 0.7227581060440565, 'reward_max': 0.7566606735186716}\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 4, 'lr': 0.0206, 'nm': 2.0, 'eps': 0.3944, 'val_loss': 0.4722, 'reward': 0.6488}: 100%|██████████| 9/9 [01:55<00:00, 12.87s/it]\n",
            "  0%|          | 0/9 [00:00<?, ?it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 4, 'reward_mean': 0.7219940240767189, 'reward_max': 0.762824049763726}\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 5, 'lr': 0.0059, 'nm': 3.8, 'eps': 0.1885, 'val_loss': 0.3795, 'reward': 0.7562}: 100%|██████████| 9/9 [01:57<00:00, 13.10s/it]\n",
            "  0%|          | 0/9 [00:00<?, ?it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 5, 'reward_mean': 0.7435776467061885, 'reward_max': 0.767639519980928}\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 6, 'lr': 0.0304, 'nm': 3.5, 'eps': 0.2047, 'val_loss': 1.1261, 'reward': 0.5696}: 100%|██████████| 9/9 [01:58<00:00, 13.21s/it]\n",
            "  0%|          | 0/9 [00:00<?, ?it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 6, 'reward_mean': 0.7266806992251911, 'reward_max': 0.771222828081519}\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 7, 'lr': 0.0353, 'nm': 4.1, 'eps': 0.176, 'val_loss': 0.8846, 'reward': 0.6258}: 100%|██████████| 9/9 [01:56<00:00, 12.95s/it]\n",
            "  0%|          | 0/9 [00:00<?, ?it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 7, 'reward_mean': 0.7240482672613099, 'reward_max': 0.7710215247587959}\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 8, 'lr': 0.0108, 'nm': 4.4, 'eps': 0.166, 'val_loss': 0.4951, 'reward': 0.7283}: 100%|██████████| 9/9 [01:59<00:00, 13.23s/it]\n",
            "  0%|          | 0/9 [00:00<?, ?it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 8, 'reward_mean': 0.7528602811508066, 'reward_max': 0.7694941854403018}\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 9, 'lr': 0.0402, 'nm': 4.4, 'eps': 0.166, 'val_loss': 0.5299, 'reward': 0.7179}: 100%|██████████| 9/9 [02:04<00:00, 13.80s/it]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "{'gen_num': 9, 'reward_mean': 0.7406596381352251, 'reward_max': 0.7712858056903273}\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9tp5dCA-hVRa",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "fa388688-333a-4d26-9c72-f9e19fa8f52a"
      },
      "source": [
        "rl_progress = run_reinforcement_learning_optimization()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "100%|██████████| 10/10 [02:57<00:00, 17.77s/it]\n",
            "100%|██████████| 10/10 [02:21<00:00, 14.12s/it]\n",
            "100%|██████████| 10/10 [02:13<00:00, 13.33s/it]\n",
            "100%|██████████| 10/10 [02:17<00:00, 13.75s/it]\n",
            "100%|██████████| 10/10 [02:17<00:00, 13.71s/it]\n",
            "100%|██████████| 10/10 [02:12<00:00, 13.21s/it]\n",
            "100%|██████████| 10/10 [02:11<00:00, 13.14s/it]\n",
            "100%|██████████| 10/10 [02:13<00:00, 13.32s/it]\n",
            "100%|██████████| 10/10 [02:14<00:00, 13.48s/it]\n",
            "100%|██████████| 10/10 [02:12<00:00, 13.25s/it]\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Eo3X-0MDhhGS"
      },
      "source": [
        "## Evaluating the Results:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-9Y1uQ4blYL1"
      },
      "source": [
        "eo_progress_ext = np.concatenate(eo_progress, 0)\n",
        "rl_progress_ext = np.concatenate(rl_progress, 0)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gTmx-7j4hX2b",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d4d5dd8a-9fd6-4663-e2c6-211326efb182"
      },
      "source": [
        "print(\"Maximum Reward Achieved by Each Algorithm:\")\n",
        "algorithms = [gs_progress, bo_progress, eo_progress_ext, rl_progress_ext]\n",
        "max_rewards_index = [np.argmax(i.T[-1]) for i in algorithms]\n",
        "df = np.stack([i[index] for i, index in zip(algorithms, max_rewards_index)])\n",
        "df = np.concatenate(([['Grid Search', 'Bayesian Optimization', 'Evolutionary Algorithms', 'Reinforcement Learning']], df.T))\n",
        "max_rewards = pd.DataFrame(df.T)\n",
        "max_rewards.columns = 'algorithms, nm, lr, eps, train_loss, val_loss, train_acc, val_acc, reward'.split(', ')\n",
        "max_rewards.set_index(\"algorithms\", inplace = True)\n",
        "pd.set_option(\"display.max_rows\", None, \"display.max_columns\", None)\n",
        "print(max_rewards)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Maximum Reward Achieved by Each Algorithm:\n",
            "                                        nm                     lr  \\\n",
            "algorithms                                                          \n",
            "Grid Search              4.666666666666666                  0.001   \n",
            "Bayesian Optimization    4.800096455988798  0.0017940862990946287   \n",
            "Evolutionary Algorithms  4.208333333333333  0.0026333333333333334   \n",
            "Reinforcement Learning   4.666666666666666                  0.001   \n",
            "\n",
            "                                         eps           train_loss  \\\n",
            "algorithms                                                          \n",
            "Grid Search              0.15877983425643916  0.39874432507902385   \n",
            "Bayesian Optimization    0.15562559348348476   0.3456091733835638   \n",
            "Evolutionary Algorithms  0.17211126028152388  0.36756317228544505   \n",
            "Reinforcement Learning   0.15877983425643916  0.38918179236352446   \n",
            "\n",
            "                                    val_loss  train_acc    val_acc  \\\n",
            "algorithms                                                           \n",
            "Grid Search               0.3876035475730896     0.8825  0.8765625   \n",
            "Bayesian Optimization     0.3360342921689153     0.8875  0.8871875   \n",
            "Evolutionary Algorithms  0.35569526420906183   0.889375    0.89125   \n",
            "Reinforcement Learning   0.38243263080716133  0.8840625    0.88375   \n",
            "\n",
            "                                     reward  \n",
            "algorithms                                   \n",
            "Grid Search              0.7659327675137778  \n",
            "Bayesian Optimization    0.7852390887704337  \n",
            "Evolutionary Algorithms  0.7712858056903273  \n",
            "Reinforcement Learning   0.7676920144517156  \n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "faE6E_tIS-nj"
      },
      "source": [
        "from matplotlib import pyplot as plt"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 297
        },
        "id": "e6EVxwRKVjM2",
        "outputId": "c70e1110-30c8-4003-bc2d-e94a2feec64d"
      },
      "source": [
        "plt.cla()\n",
        "plt.bar(np.arange(4), height=[float(i) for i in df[-1]], width=0.75)\n",
        "plt.ylim([0.764, 0.79])\n",
        "plt.ylabel(\"Number of Experiments\")\n",
        "plt.xticks(np.arange(4), [\"Grid\\nSearch\", \"Bayesian\\nOptimization\", \"Evolutionary\\nOptimization\", \"Reinforcement\\nLearning\"])\n",
        "plt.xlabel('''\n",
        "\n",
        "Best Reward Achieved''', fontsize=12)\n",
        "plt.tight_layout()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 297
        },
        "id": "kmWJO6asX7e1",
        "outputId": "c1784e31-2c0d-4874-f6fc-d4f5f475be69"
      },
      "source": [
        "best_grid_search_reward_index = np.argmax(gs_progress.T[-1])\n",
        "best_grid_search_reward = gs_progress[best_grid_search_reward_index][-1]\n",
        "\n",
        "timely_indices = []\n",
        "for algorithm in algorithms:\n",
        "  for index, row in enumerate(algorithm):\n",
        "    if row[-1]>=best_grid_search_reward:\n",
        "      timely_indices.append(index)\n",
        "      break\n",
        "\n",
        "plt.cla()\n",
        "plt.bar(np.arange(4), height=timely_indices, width=0.75)\n",
        "plt.ylabel(\"Number of Experiments\")\n",
        "plt.xticks(np.arange(4), [\"Grid\\nSearch\", \"Bayesian\\nOptimization\", \"Evolutionary\\nOptimization\", \"Reinforcement\\nLearning\"])\n",
        "plt.xlabel('''\n",
        "\n",
        "Number of experiments run by different \n",
        "methodologies to achieve reward greater \n",
        "than or equal to baseline (Grid Search) rewards.''', fontsize=12)\n",
        "plt.tight_layout()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VGU2ymfUuFoS"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}