PIN_recognition.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "toc_visible": true,
      "mount_file_id": "1VHMJd8VrEegRdNX8tXyexZpTzxEY6sF1",
      "authorship_tag": "ABX9TyNXYLUXAzFndN78Fj+0kCWC",
      "include_colab_link": true
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "gpuClass": "standard"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/matteonerini/pin-side-channel-attacks/blob/main/PIN_recognition.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "N4Gh6Ukcrico"
      },
      "source": [
        "# Digit Classification with SVM, RF, k-NN, and MLP"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TWqZf1NYb87q"
      },
      "source": [
        "## Data Pre-processing\n",
        "\n",
        "The dataset is uploaded from a .txt file and pre-processed."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "amSPm9cerIFi"
      },
      "source": [
        "import pandas as pd\n",
        "import numpy as np\n",
        "from sklearn.preprocessing import StandardScaler\n",
        "\n",
        "short = True\n",
        "url = 'https://raw.githubusercontent.com/matteonerini/pin-side-channel-attacks/main/dataset.txt'\n",
        "\n",
        "names = ['Digit','Acc_X','Acc_Y','Acc_Z','Gir_X','Gir_Y','Gir_Z','Gra_X','Gra_Y','Gra_Z',\n",
        "         'Lin_X','Lin_Y','Lin_Z','Rot_Z','Ori_X','Ori_Y','M']\n",
        "\n",
        "dataset = pd.read_csv(url, names=names).values\n",
        "\n",
        "n_class = 10\n",
        "n_entries, n_feat = dataset.shape\n",
        "n_feat = n_feat -1\n",
        "test_size = 0.1 if short else 0.05\n",
        "n_test_digit = int(n_entries*test_size/n_class)\n",
        "\n",
        "X = dataset[:,1:17]\n",
        "y = dataset[:,0]\n",
        "\n",
        "# Feature scaling\n",
        "scaler=StandardScaler()\n",
        "X_scaled = scaler.fit_transform(X)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XPj8YwVmdFHg"
      },
      "source": [
        "## Model Generation, Training and Testing\n",
        "\n",
        "The dataset is randomly divided in train and test sets. The model is generated, trained and tested with different seeds."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "h5QX5UVtlmTu"
      },
      "source": [
        "from datetime import datetime\n",
        "import csv\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn import svm\n",
        "from sklearn.ensemble import RandomForestClassifier\n",
        "from sklearn.neighbors import KNeighborsClassifier\n",
        "from sklearn.neural_network import MLPClassifier\n",
        "from sklearn.metrics import confusion_matrix\n",
        "import heapq\n",
        "from itertools import combinations\n",
        "import scipy.special\n",
        "\n",
        "model = 'rf' # 'svm', 'rf', 'knn', or 'mlp'\n",
        "divType = 'prod' # 'prod' for {'svm','rf','mlp'}, or 'sum' for 'knn'\n",
        "n_div = 10\n",
        "n_seeds = 20"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "YWjR4wVnmmlm"
      },
      "source": [
        "In the following cell, the diversity is implemented as follows:\n",
        "\n",
        "1. The probabilities are calculated for each class.\n",
        "2. The $d$ probabilities obtained for each class are multiplied/summed up, depending on the variable `divType`.\n",
        "3. The class with the gratest product/sum is selected."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RRRr_LotLdjK"
      },
      "source": [
        "csvFileName = \"./\" + model + '-' + divType + '.csv'\n",
        "\n",
        "# Precompute all the combinations with iDiv items and the combination indexes\n",
        "comb_list_all = []\n",
        "iComb_list_all = []\n",
        "for div in range(2,n_div+1):\n",
        "    comb_list_all.append(list(combinations(range(0,n_test_digit), div)))\n",
        "    iComb_list_all.append(np.array(range(int(scipy.special.binom(n_test_digit, div)))))\n",
        "\n",
        "for seed in range(n_seeds):\n",
        "\n",
        "    # Split into train test sets with seeds from 1 to n_seeds\n",
        "    X_train, X_test, y_train, y_test = train_test_split(X_scaled, y, test_size=test_size, random_state=seed+1, stratify=y)\n",
        "    \n",
        "    if model == 'svm':\n",
        "        # SVM\n",
        "        print('Model: ' + model + '. Computing seed: ' + str(seed+1))\n",
        "        clf = svm.SVC(C=3, decision_function_shape='ovr', probability=True)\n",
        "        clf.fit(X_train, y_train)\n",
        "    elif model == 'rf':\n",
        "        # RF\n",
        "        print('Model: ' + model + '. Computing seed: ' + str(seed+1))\n",
        "        clf = RandomForestClassifier(n_estimators=150, min_samples_split = 4, min_samples_leaf = 2)\n",
        "        clf.fit(X_train, y_train)\n",
        "    elif model == 'knn':\n",
        "        # k-NN\n",
        "        print('Model: ' + model + '. Computing seed: ' + str(seed+1))\n",
        "        clf = KNeighborsClassifier(n_neighbors=8)\n",
        "        clf.fit(X_train, y_train)\n",
        "    elif model == 'mlp':\n",
        "        # MLP\n",
        "        print('Model: ' + model + '. Computing seed: ' + str(seed+1))\n",
        "        clf = MLPClassifier(hidden_layer_sizes=(100,50), max_iter=1000, learning_rate_init=5e-3, early_stopping=True,\n",
        "                            n_iter_no_change=25)\n",
        "        clf.fit(X_train, y_train)\n",
        "    else:\n",
        "        print('Insert a correct model name')\n",
        "\n",
        "    # div==1 (no diversity)\n",
        "    y_pred = clf.predict(X_test)\n",
        "    conf1 = confusion_matrix(y_test, y_pred)\n",
        "    with open(csvFileName, 'a', newline='') as file:\n",
        "        writer = csv.writer(file)\n",
        "        writer.writerow(conf1.flatten())\n",
        "    y_proba = clf.predict_proba(X_test)\n",
        "    #y_proba_max = []\n",
        "    #for i in range(len(y_test)):\n",
        "    #    index_i = np.argmax(y_proba[i,:])\n",
        "    #    y_proba_max.append(index_i+1)\n",
        "    y_proba_max2 = []\n",
        "    for i in range(len(y_test)):\n",
        "        index_temp2 = heapq.nlargest(2, range(len(y_proba[i,:])), key=y_proba[i,:].__getitem__)\n",
        "        index_i2 = index_temp2[1]\n",
        "        y_proba_max2.append(index_i2+1)\n",
        "    conf2 = confusion_matrix(y_test, y_proba_max2)\n",
        "    with open(csvFileName, 'a', newline='') as file:\n",
        "        writer = csv.writer(file)\n",
        "        writer.writerow(conf2.flatten())\n",
        "\n",
        "    # div!=1 (diversity)\n",
        "    for div in range(2,n_div+1):\n",
        "\n",
        "        # The testset is ordered\n",
        "        testset = np.column_stack((y_test, X_test))\n",
        "        testset_ord = testset[testset[:,0].argsort()]\n",
        "\n",
        "        comb_list = comb_list_all[div-2]\n",
        "        iComb_list = iComb_list_all[div-2]\n",
        "        \n",
        "        conf_temp1 = []\n",
        "        conf_temp2 = []\n",
        "        for j in range(n_class):\n",
        "    \n",
        "            testsetj = testset_ord[j*n_test_digit:(j+1)*n_test_digit,:]\n",
        "            y_proba_div_temp = []\n",
        "\n",
        "            np.random.seed(seed+1)\n",
        "            np.random.shuffle(testsetj)\n",
        "    \n",
        "            # List all combinations with div items\n",
        "            #comb_list = list(combinations(range(0,n_test_digit), div))\n",
        "            #iComb_list = np.array(range(int(scipy.special.binom(n_test_digit, div))))\n",
        "            #np.random.shuffle(iComb_list)\n",
        "            np.random.shuffle(iComb_list)\n",
        "\n",
        "            for iComb in range(min(len(iComb_list), 250)):\n",
        "                i = comb_list[iComb_list[iComb]]\n",
        "                if div == 2: y_proba_div_comb = clf.predict_proba(np.vstack((testsetj[i[0],1:17], testsetj[i[1],1:17])))\n",
        "                if div == 3: y_proba_div_comb = clf.predict_proba(np.vstack((testsetj[i[0],1:17], testsetj[i[1],1:17],\n",
        "                                                                             testsetj[i[2],1:17])))\n",
        "                if div == 4: y_proba_div_comb = clf.predict_proba(np.vstack((testsetj[i[0],1:17], testsetj[i[1],1:17],\n",
        "                                                                             testsetj[i[2],1:17], testsetj[i[3],1:17])))\n",
        "                if div == 5: y_proba_div_comb = clf.predict_proba(np.vstack((testsetj[i[0],1:17], testsetj[i[1],1:17],\n",
        "                                                                             testsetj[i[2],1:17], testsetj[i[3],1:17],\n",
        "                                                                             testsetj[i[4],1:17])))\n",
        "                if div == 6: y_proba_div_comb = clf.predict_proba(np.vstack((testsetj[i[0],1:17], testsetj[i[1],1:17],\n",
        "                                                                             testsetj[i[2],1:17], testsetj[i[3],1:17],\n",
        "                                                                             testsetj[i[4],1:17], testsetj[i[5],1:17])))\n",
        "                if div == 7: y_proba_div_comb = clf.predict_proba(np.vstack((testsetj[i[0],1:17], testsetj[i[1],1:17],\n",
        "                                                                             testsetj[i[2],1:17], testsetj[i[3],1:17],\n",
        "                                                                             testsetj[i[4],1:17], testsetj[i[5],1:17],\n",
        "                                                                             testsetj[i[6],1:17])))\n",
        "                if div == 8: y_proba_div_comb = clf.predict_proba(np.vstack((testsetj[i[0],1:17], testsetj[i[1],1:17],\n",
        "                                                                             testsetj[i[2],1:17], testsetj[i[3],1:17],\n",
        "                                                                             testsetj[i[4],1:17], testsetj[i[5],1:17],\n",
        "                                                                             testsetj[i[6],1:17], testsetj[i[7],1:17])))\n",
        "                if div == 9: y_proba_div_comb = clf.predict_proba(np.vstack((testsetj[i[0],1:17], testsetj[i[1],1:17],\n",
        "                                                                             testsetj[i[2],1:17], testsetj[i[3],1:17],\n",
        "                                                                             testsetj[i[4],1:17], testsetj[i[5],1:17],\n",
        "                                                                             testsetj[i[6],1:17], testsetj[i[7],1:17],\n",
        "                                                                             testsetj[i[8],1:17])))\n",
        "                if div == 10: y_proba_div_comb = clf.predict_proba(np.vstack((testsetj[i[0],1:17], testsetj[i[1],1:17],\n",
        "                                                                             testsetj[i[2],1:17], testsetj[i[3],1:17],\n",
        "                                                                             testsetj[i[4],1:17], testsetj[i[5],1:17],\n",
        "                                                                             testsetj[i[6],1:17], testsetj[i[7],1:17],\n",
        "                                                                             testsetj[i[8],1:17], testsetj[i[9],1:17])))\n",
        "                if divType == 'prod':\n",
        "                    y_proba_div_temp.append(np.prod(y_proba_div_comb, axis = 0))\n",
        "                elif divType == 'sum':\n",
        "                    y_proba_div_temp.append(np.sum(y_proba_div_comb, axis = 0))\n",
        "                else:\n",
        "                    print('Insert a correct divType name')\n",
        "            y_proba_div = np.array(y_proba_div_temp)\n",
        "    \n",
        "            y_proba_max_divj = []\n",
        "            y_proba_max_divj2 = []\n",
        "            for i in range(len(y_proba_div)):\n",
        "                index_i = np.argmax(y_proba_div[i,:])\n",
        "                y_proba_max_divj.append(index_i+1)\n",
        "                index_temp2 = heapq.nlargest(2, range(len(y_proba_div[i,:])), key=y_proba_div[i,:].__getitem__)\n",
        "                index_i2 = index_temp2[1]\n",
        "                y_proba_max_divj2.append(index_i2+1)\n",
        "\n",
        "            y_proba_max_divj_conf = []\n",
        "            for i in range(n_class):\n",
        "                y_proba_max_divj_conf.append(np.count_nonzero(np.array(y_proba_max_divj) == i+1))\n",
        "            conf_temp1.append(y_proba_max_divj_conf)\n",
        "\n",
        "            y_proba_max_divj_conf2 = []\n",
        "            for i in range(n_class):\n",
        "                y_proba_max_divj_conf2.append(np.count_nonzero(np.array(y_proba_max_divj2) == i+1))\n",
        "            conf_temp2.append(y_proba_max_divj_conf2)\n",
        "\n",
        "        conf1 = np.array(conf_temp1)\n",
        "        with open(csvFileName, 'a', newline='') as file:\n",
        "            writer = csv.writer(file)\n",
        "            writer.writerow(conf1.flatten())\n",
        "        conf2 = np.array(conf_temp2)\n",
        "        with open(csvFileName, 'a', newline='') as file:\n",
        "            writer = csv.writer(file)\n",
        "            writer.writerow(conf2.flatten())"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1HCSyH2xVlok"
      },
      "source": [
        "## Plot Performance\n",
        "\n",
        "These cells read the .csv file and plot the data. The .csv file is formatted in the following way:\n",
        "\n",
        "```\n",
        "1        xx ... x <- div = 1, conf1, seed = 1\n",
        "2        xx ... x <- div = 1, conf2\n",
        "3        xx ... x <- div = 2, conf1\n",
        "4        xx ... x <- div = 2, conf2\n",
        "...\n",
        "2n_div-1 xx ... x <- div = 8, conf1\n",
        "2n_div   xx ... x <- div = 8, conf2\n",
        "2n_div+1 xx ... x <- div = 1, conf1, seed = 2\n",
        "2n_div+2 xx ... x <- div = 1, conf2\n",
        "```\n",
        "\n",
        "where `xx ... x` is a 100 elements confusion matrix, `conf1` means at the first attempt, and `conf2` means at the second attempt."
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Plot the confusion matrix of one predictor for div=1 and div=10\n",
        "\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "import seaborn as sn\n",
        "from google.colab import widgets\n",
        "\n",
        "n_div = 10\n",
        "n_seeds = 20\n",
        "iDiv = [1, 10]\n",
        "\n",
        "dataset = pd.read_csv('rf-prod.csv', header=None).values\n",
        "#dataset = pd.read_csv('svm-prod.csv', header=None).values\n",
        "#dataset = pd.read_csv('knn-sum.csv', header=None).values\n",
        "#dataset = pd.read_csv('mlp-prod.csv', header=None).values\n",
        "\n",
        "perf_all1 = []\n",
        "perf_all2 = []\n",
        "for iRow in range(len(dataset)):\n",
        "    mat = np.reshape(dataset[iRow,:], (10, 10))\n",
        "    mat = 100*mat / (mat[0].sum())\n",
        "    if iRow % 2 == 0:\n",
        "        perf_all1.append(mat)\n",
        "    else:\n",
        "        perf_all2.append(mat)\n",
        "\n",
        "perf1 = [0] * n_div\n",
        "for i in range(len(perf_all1)):\n",
        "    perf1[i%n_div] = perf1[i%n_div] + perf_all1[i]    \n",
        "perf1 = np.array(perf1) / n_seeds\n",
        "\n",
        "tb = widgets.TabBar(['Div '+str(i) for i in iDiv])\n",
        "\n",
        "for i in iDiv:\n",
        "    with tb.output_to('Div '+str(i)):\n",
        "        mat_cm = pd.DataFrame(np.round(perf1[i-1]), (1,2,3,4,5,6,7,8,9,0), (1,2,3,4,5,6,7,8,9,0))\n",
        "        fig, ax = plt.subplots()\n",
        "        sn.heatmap(mat_cm, annot=True, cbar=False, cmap='Blues')\n",
        "        ax.set_ylabel('True Class')\n",
        "        ax.set_xlabel('Predicted Class')\n",
        "        plt.yticks(rotation=0)\n",
        "        plt.savefig('knn-conf'+str(i)+'.pdf', bbox_inches='tight')"
      ],
      "metadata": {
        "colab": {
          "resources": {
            "http://localhost:8080/nbextensions/google.colab/tabbar.css": {
              "data": "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",
              "ok": true,
              "headers": [
                [
                  "content-type",
                  "text/css"
                ]
              ],
              "status": 200,
              "status_text": ""
            },
            "http://localhost:8080/nbextensions/google.colab/tabbar_main.min.js": {
              "data": "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",
              "ok": true,
              "headers": [
                [
                  "content-type",
                  "application/javascript"
                ]
              ],
              "status": 200,
              "status_text": ""
            }
          },
          "base_uri": "https://localhost:8080/",
          "height": 320
        },
        "id": "QFfuKDg2reSV",
        "outputId": "a5cefa28-618c-45a3-8170-6aa696f33c6a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ],
            "text/html": [
              "<link rel=stylesheet type=text/css href='/nbextensions/google.colab/tabbar.css'></link>"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ],
            "text/html": [
              "<script src='/nbextensions/google.colab/tabbar_main.min.js'></script>"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ],
            "text/html": [
              "<div id=\"id2\"></div>"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.Javascript object>"
            ],
            "application/javascript": [
              "window[\"ced4f9b2-c735-11ed-9e6b-0242ac1c000c\"] = colab_lib.createTabBar({\"location\": \"top\", \"elementId\": \"id2\", \"tabNames\": [\"Div 1\", \"Div 10\"], \"initialSelection\": 0, \"contentBorder\": [\"0px\"], \"contentHeight\": [\"initial\"], \"borderColor\": [\"#a7a7a7\"]});\n",
              "//# sourceURL=js_45defa8f78"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.Javascript object>"
            ],
            "application/javascript": [
              "window[\"ced5d922-c735-11ed-9e6b-0242ac1c000c\"] = window[\"id2\"].setSelectedTabIndex(0);\n",
              "//# sourceURL=js_3ea2cb7774"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.Javascript object>"
            ],
            "application/javascript": [
              "window[\"ceda82ce-c735-11ed-9e6b-0242ac1c000c\"] = google.colab.output.getActiveOutputArea();\n",
              "//# sourceURL=js_49002a43ea"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.Javascript object>"
            ],
            "application/javascript": [
              "window[\"cedb0c80-c735-11ed-9e6b-0242ac1c000c\"] = document.querySelector(\"#id2_content_0\");\n",
              "//# sourceURL=js_130508f879"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.Javascript object>"
            ],
            "application/javascript": [
              "window[\"cedc61c0-c735-11ed-9e6b-0242ac1c000c\"] = google.colab.output.setActiveOutputArea(window[\"cedb0c80-c735-11ed-9e6b-0242ac1c000c\"]);\n",
              "//# sourceURL=js_a45d3f0686"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.Javascript object>"
            ],
            "application/javascript": [
              "window[\"cedcfc7a-c735-11ed-9e6b-0242ac1c000c\"] = window[\"id2\"].setSelectedTabIndex(0);\n",
              "//# sourceURL=js_49f55aea46"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.Javascript object>"
            ],
            "application/javascript": [
              "window[\"cf93543e-c735-11ed-9e6b-0242ac1c000c\"] = google.colab.output.setActiveOutputArea(window[\"ceda82ce-c735-11ed-9e6b-0242ac1c000c\"]);\n",
              "//# sourceURL=js_3f642579d2"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.Javascript object>"
            ],
            "application/javascript": [
              "window[\"cf95bb7a-c735-11ed-9e6b-0242ac1c000c\"] = google.colab.output.getActiveOutputArea();\n",
              "//# sourceURL=js_e84127ee48"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.Javascript object>"
            ],
            "application/javascript": [
              "window[\"cf95faf4-c735-11ed-9e6b-0242ac1c000c\"] = document.querySelector(\"#id2_content_1\");\n",
              "//# sourceURL=js_2cdf85f87b"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.Javascript object>"
            ],
            "application/javascript": [
              "window[\"cf96738a-c735-11ed-9e6b-0242ac1c000c\"] = google.colab.output.setActiveOutputArea(window[\"cf95faf4-c735-11ed-9e6b-0242ac1c000c\"]);\n",
              "//# sourceURL=js_13895e410c"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.Javascript object>"
            ],
            "application/javascript": [
              "window[\"cf96d398-c735-11ed-9e6b-0242ac1c000c\"] = window[\"id2\"].setSelectedTabIndex(1);\n",
              "//# sourceURL=js_5d9f1ab721"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.Javascript object>"
            ],
            "application/javascript": [
              "window[\"d0080374-c735-11ed-9e6b-0242ac1c000c\"] = google.colab.output.setActiveOutputArea(window[\"cf95bb7a-c735-11ed-9e6b-0242ac1c000c\"]);\n",
              "//# sourceURL=js_d7f477ba5f"
            ]
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 297
        },
        "id": "K7M3BjA0I50g",
        "outputId": "43af5d2e-fec2-41f9-d759-ac424359b7a0"
      },
      "source": [
        "# Plot accuracy of all predictors (1st and 2nd) for different diversity\n",
        "\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "n_div = 10\n",
        "n_seeds = 20\n",
        "\n",
        "fileNameMLP = 'mlp-prod.csv'\n",
        "fileNameRF  = 'rf-prod.csv'\n",
        "fileNameKNN = 'knn-sum.csv'\n",
        "fileNameSVM = 'svm-prod.csv'\n",
        "\n",
        "# RF\n",
        "dataset = pd.read_csv(fileNameRF, header=None).values\n",
        "\n",
        "perf_all_rf1 = []\n",
        "perf_all_rf2 = []\n",
        "for iRow in range(len(dataset)):\n",
        "    mat = np.reshape(dataset[iRow,:], (10, 10))\n",
        "    if iRow % 2 == 0:\n",
        "        perf_all_rf1.append(100*np.trace(mat)/(mat[0].sum()*10))\n",
        "    else:\n",
        "        perf_all_rf2.append(100*np.trace(mat)/(mat[0].sum()*10))\n",
        "\n",
        "perf_rf1 = [0] * n_div\n",
        "perf_rf2 = [0] * n_div\n",
        "for i in range(len(perf_all_rf1)):\n",
        "    perf_rf1[i%n_div] = perf_rf1[i%n_div] + perf_all_rf1[i]    \n",
        "    perf_rf2[i%n_div] = perf_rf2[i%n_div] + perf_all_rf2[i]\n",
        "perf_rf1 = np.array(perf_rf1) / n_seeds\n",
        "perf_rf2 = np.array(perf_rf2) / n_seeds\n",
        "\n",
        "# SVM\n",
        "dataset = pd.read_csv(fileNameSVM, header=None).values\n",
        "\n",
        "perf_all_svm1 = []\n",
        "perf_all_svm2 = []\n",
        "for iRow in range(len(dataset)):\n",
        "    mat = np.reshape(dataset[iRow,:], (10, 10))\n",
        "    if iRow % 2 == 0:\n",
        "        perf_all_svm1.append(100*np.trace(mat)/(mat[0].sum()*10))\n",
        "    else:\n",
        "        perf_all_svm2.append(100*np.trace(mat)/(mat[0].sum()*10))\n",
        "\n",
        "perf_svm1 = [0] * n_div\n",
        "perf_svm2 = [0] * n_div\n",
        "for i in range(len(perf_all_svm1)):\n",
        "    perf_svm1[i%n_div] = perf_svm1[i%n_div] + perf_all_svm1[i]    \n",
        "    perf_svm2[i%n_div] = perf_svm2[i%n_div] + perf_all_svm2[i]\n",
        "perf_svm1 = np.array(perf_svm1) / n_seeds\n",
        "perf_svm2 = np.array(perf_svm2) / n_seeds\n",
        "\n",
        "# MLP\n",
        "dataset = pd.read_csv(fileNameMLP, header=None).values\n",
        "\n",
        "perf_all_mlp1 = []\n",
        "perf_all_mlp2 = []\n",
        "for iRow in range(len(dataset)):\n",
        "    mat = np.reshape(dataset[iRow,:], (10, 10))\n",
        "    if iRow % 2 == 0:\n",
        "        perf_all_mlp1.append(100*np.trace(mat)/(mat[0].sum()*10))\n",
        "    else:\n",
        "        perf_all_mlp2.append(100*np.trace(mat)/(mat[0].sum()*10))\n",
        "\n",
        "perf_mlp1 = [0] * n_div\n",
        "perf_mlp2 = [0] * n_div\n",
        "for i in range(len(perf_all_mlp1)):\n",
        "    perf_mlp1[i%n_div] = perf_mlp1[i%n_div] + perf_all_mlp1[i]    \n",
        "    perf_mlp2[i%n_div] = perf_mlp2[i%n_div] + perf_all_mlp2[i]\n",
        "perf_mlp1 = np.array(perf_mlp1) / n_seeds\n",
        "perf_mlp2 = np.array(perf_mlp2) / n_seeds\n",
        "\n",
        "# k-NN\n",
        "dataset = pd.read_csv(fileNameKNN, header=None).values\n",
        "\n",
        "perf_all_knn1 = []\n",
        "perf_all_knn2 = []\n",
        "for iRow in range(len(dataset)):\n",
        "    mat = np.reshape(dataset[iRow,:], (10, 10))\n",
        "    if iRow % 2 == 0:\n",
        "        perf_all_knn1.append(100*np.trace(mat)/(mat[0].sum()*10))\n",
        "    else:\n",
        "        perf_all_knn2.append(100*np.trace(mat)/(mat[0].sum()*10))\n",
        "\n",
        "perf_knn1 = [0] * n_div\n",
        "perf_knn2 = [0] * n_div\n",
        "for i in range(len(perf_all_knn1)):\n",
        "    perf_knn1[i%n_div] = perf_knn1[i%n_div] + perf_all_knn1[i]    \n",
        "    perf_knn2[i%n_div] = perf_knn2[i%n_div] + perf_all_knn2[i]\n",
        "perf_knn1 = np.array(perf_knn1) / n_seeds\n",
        "perf_knn2 = np.array(perf_knn2) / n_seeds\n",
        "\n",
        "# Single digit plot\n",
        "labels = list(range(1,n_div+1))\n",
        "x = np.arange(len(labels))  # the label locations\n",
        "width = 0.2  # the width of the bars\n",
        "fig, ax = plt.subplots()\n",
        "ax.grid(axis='y', zorder=0)\n",
        "rects1 = ax.bar(x - 3*width/2, perf_rf1 , width, label='$A_1$ RF', zorder=3, color = '#E74C3C')\n",
        "rects3 = ax.bar(x - width/2  , perf_svm1, width, label='$A_1$ SVM', zorder=3, color = '#3498DB')\n",
        "rects5 = ax.bar(x + width/2  , perf_mlp1, width, label='$A_1$ NN', zorder=3, color = '#F39C12')\n",
        "rects7 = ax.bar(x + 3*width/2, perf_knn1, width, label='$A_1$ K-NN', zorder=3, color = '#27AE60')\n",
        "rects2 = ax.bar(x - 3*width/2, perf_rf2 , width, bottom=perf_rf1, label='$A_2$ RF', zorder=3, color = '#F5B7B1')\n",
        "rects4 = ax.bar(x - width/2  , perf_svm2, width, bottom=perf_svm1, label='$A_2$ SVM', zorder=3, color = '#AED6F1')\n",
        "rects6 = ax.bar(x + width/2  , perf_mlp2, width, bottom=perf_mlp1, label='$A_2$ NN', zorder=3, color = '#FAD7A0')\n",
        "rects8 = ax.bar(x + 3*width/2, perf_knn2, width, bottom=perf_knn1, label='$A_2$ K-NN', zorder=3, color = '#A9DFBF')\n",
        "\n",
        "ax.set_xlabel('Diversity')\n",
        "ax.set_ylabel('Accuracy [%]')\n",
        "ax.set_xticks(x)\n",
        "ax.set_xticklabels(labels)\n",
        "ax.set_ylim((0, 100))\n",
        "ax.legend(labelspacing=0.25, ncol=2, loc=4, framealpha=0.95)\n",
        "fig.tight_layout()\n",
        "plt.savefig('all-digit-acc.pdf', bbox_inches='tight')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 297
        },
        "id": "iDsyGabOLVPw",
        "outputId": "57f30cd0-064e-47b1-d678-7b6943f75446"
      },
      "source": [
        "# Plot F1-score of all predictors (1st and 2nd) for different diversity\n",
        "\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "n_div = 10\n",
        "n_seeds = 20\n",
        "\n",
        "fileNameMLP = 'mlp-prod.csv'\n",
        "fileNameRF  = 'rf-prod.csv'\n",
        "fileNameKNN = 'knn-sum.csv'\n",
        "fileNameSVM = 'svm-prod.csv'\n",
        "\n",
        "# RF\n",
        "dataset = pd.read_csv(fileNameRF, header=None).values\n",
        "\n",
        "perf_all_rf1 = []\n",
        "perf_all_rf2 = []\n",
        "for iRow in range(len(dataset)):\n",
        "    mat = np.reshape(dataset[iRow,:], (10, 10))\n",
        "    F1_k = []\n",
        "    for ik in range(10):\n",
        "        F1_k.append(2*mat[ik,ik] / (2*mat[ik,ik] + (np.sum(mat[:,ik])-mat[ik,ik]) + (np.sum(mat[ik,:])-mat[ik,ik])))\n",
        "    if iRow % 2 == 0:\n",
        "        perf_all_rf1.append(100*np.mean(F1_k))\n",
        "    else:\n",
        "        perf_all_rf2.append(100*np.mean(F1_k))\n",
        "\n",
        "perf_rf1 = [0] * n_div\n",
        "perf_rf2 = [0] * n_div\n",
        "for i in range(len(perf_all_rf1)):\n",
        "    perf_rf1[i%n_div] = perf_rf1[i%n_div] + perf_all_rf1[i]    \n",
        "    perf_rf2[i%n_div] = perf_rf2[i%n_div] + perf_all_rf2[i]\n",
        "perf_rf1 = np.array(perf_rf1) / n_seeds\n",
        "perf_rf2 = np.array(perf_rf2) / n_seeds\n",
        "\n",
        "# SVM\n",
        "dataset = pd.read_csv(fileNameSVM, header=None).values\n",
        "\n",
        "perf_all_svm1 = []\n",
        "perf_all_svm2 = []\n",
        "for iRow in range(len(dataset)):\n",
        "    mat = np.reshape(dataset[iRow,:], (10, 10))\n",
        "    F1_k = []\n",
        "    for ik in range(10):\n",
        "        F1_k.append(2*mat[ik,ik] / (2*mat[ik,ik] + (np.sum(mat[:,ik])-mat[ik,ik]) + (np.sum(mat[ik,:])-mat[ik,ik])))\n",
        "    if iRow % 2 == 0:\n",
        "        perf_all_svm1.append(100*np.mean(F1_k))\n",
        "    else:\n",
        "        perf_all_svm2.append(100*np.mean(F1_k))\n",
        "\n",
        "perf_svm1 = [0] * n_div\n",
        "perf_svm2 = [0] * n_div\n",
        "for i in range(len(perf_all_svm1)):\n",
        "    perf_svm1[i%n_div] = perf_svm1[i%n_div] + perf_all_svm1[i]    \n",
        "    perf_svm2[i%n_div] = perf_svm2[i%n_div] + perf_all_svm2[i]\n",
        "perf_svm1 = np.array(perf_svm1) / n_seeds\n",
        "perf_svm2 = np.array(perf_svm2) / n_seeds\n",
        "\n",
        "# MLP\n",
        "dataset = pd.read_csv(fileNameMLP, header=None).values\n",
        "\n",
        "perf_all_mlp1 = []\n",
        "perf_all_mlp2 = []\n",
        "for iRow in range(len(dataset)):\n",
        "    mat = np.reshape(dataset[iRow,:], (10, 10))\n",
        "    F1_k = []\n",
        "    for ik in range(10):\n",
        "        F1_k.append(2*mat[ik,ik] / (2*mat[ik,ik] + (np.sum(mat[:,ik])-mat[ik,ik]) + (np.sum(mat[ik,:])-mat[ik,ik])))\n",
        "    if iRow % 2 == 0:\n",
        "        perf_all_mlp1.append(100*np.mean(F1_k))\n",
        "    else:\n",
        "        perf_all_mlp2.append(100*np.mean(F1_k))\n",
        "\n",
        "perf_mlp1 = [0] * n_div\n",
        "perf_mlp2 = [0] * n_div\n",
        "for i in range(len(perf_all_mlp1)):\n",
        "    perf_mlp1[i%n_div] = perf_mlp1[i%n_div] + perf_all_mlp1[i]    \n",
        "    perf_mlp2[i%n_div] = perf_mlp2[i%n_div] + perf_all_mlp2[i]\n",
        "perf_mlp1 = np.array(perf_mlp1) / n_seeds\n",
        "perf_mlp2 = np.array(perf_mlp2) / n_seeds\n",
        "\n",
        "# k-NN\n",
        "dataset = pd.read_csv(fileNameKNN, header=None).values\n",
        "\n",
        "perf_all_knn1 = []\n",
        "perf_all_knn2 = []\n",
        "for iRow in range(len(dataset)):\n",
        "    mat = np.reshape(dataset[iRow,:], (10, 10))\n",
        "    F1_k = []\n",
        "    for ik in range(10):\n",
        "        F1_k.append(2*mat[ik,ik] / (2*mat[ik,ik] + (np.sum(mat[:,ik])-mat[ik,ik]) + (np.sum(mat[ik,:])-mat[ik,ik])))\n",
        "    if iRow % 2 == 0:\n",
        "        perf_all_knn1.append(100*np.mean(F1_k))\n",
        "    else:\n",
        "        perf_all_knn2.append(100*np.mean(F1_k))\n",
        "\n",
        "perf_knn1 = [0] * n_div\n",
        "perf_knn2 = [0] * n_div\n",
        "for i in range(len(perf_all_knn1)):\n",
        "    perf_knn1[i%n_div] = perf_knn1[i%n_div] + perf_all_knn1[i]    \n",
        "    perf_knn2[i%n_div] = perf_knn2[i%n_div] + perf_all_knn2[i]\n",
        "perf_knn1 = np.array(perf_knn1) / n_seeds\n",
        "perf_knn2 = np.array(perf_knn2) / n_seeds\n",
        "\n",
        "# Single digit plot\n",
        "labels = list(range(1,n_div+1))\n",
        "x = np.arange(len(labels))  # the label locations\n",
        "width = 0.2  # the width of the bars\n",
        "fig, ax = plt.subplots()\n",
        "ax.grid(axis='y', zorder=0)\n",
        "rects1 = ax.bar(x - 3*width/2, perf_rf1 , width, label='RF', zorder=3, color = '#E74C3C')\n",
        "rects3 = ax.bar(x - width/2  , perf_svm1, width, label='SVM', zorder=3, color = '#3498DB')\n",
        "rects5 = ax.bar(x + width/2  , perf_mlp1, width, label='NN', zorder=3, color = '#F39C12')\n",
        "rects7 = ax.bar(x + 3*width/2, perf_knn1, width, label='K-NN', zorder=3, color = '#27AE60')\n",
        "\n",
        "ax.set_xlabel('Diversity')\n",
        "ax.set_ylabel('F-score [%]')\n",
        "ax.set_xticks(x)\n",
        "ax.set_xticklabels(labels)\n",
        "ax.set_ylim((0, 100))\n",
        "ax.legend(labelspacing=0.25, ncol=1, loc=4, framealpha=0.95)\n",
        "fig.tight_layout()\n",
        "plt.savefig('all-digit-F.pdf', bbox_inches='tight')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "H8mBvc274pJD"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}