federated-learning-and-bfv-scheme.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "c8e6301c",
   "metadata": {},
   "source": [
    "# Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "ba6fc3b2",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from imblearn.over_sampling import RandomOverSampler\n",
    "from torch.autograd import Variable\n",
    "import torch"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "74779994",
   "metadata": {},
   "source": [
    "### Dataset 1\n",
    "\n",
    "[source](https://www.kaggle.com/datasets/yasserh/breast-cancer-dataset)\n",
    "\n",
    "**Description:**\n",
    "Breast cancer is the most common cancer amongst women in the world. It accounts for 25% of all cancer cases, and affected over 2.1 Million people in 2015 alone. It starts when cells in the breast begin to grow out of control. These cells usually form tumors that can be seen via X-ray or felt as lumps in the breast area.\n",
    "\n",
    "The key challenges against it’s detection is how to classify tumors into malignant (cancerous) or benign(non cancerous). We ask you to complete the analysis of classifying these tumors using machine learning (with SVMs) and the Breast Cancer Wisconsin (Diagnostic) Dataset.\n",
    "\n",
    "**Acknowledgements:**\n",
    "This dataset has been referred from Kaggle.\n",
    "\n",
    "**Objective:**\n",
    "* Understand the Dataset & cleanup (if required).\n",
    "* Build classification models to predict whether the cancer type is Malignant or Benign.\n",
    "* Also fine-tune the hyperparameters & compare the evaluation metrics of various classification algorithms."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "4b3918f4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape: (1202, 32)\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>id</th>\n",
       "      <th>radius_mean</th>\n",
       "      <th>texture_mean</th>\n",
       "      <th>perimeter_mean</th>\n",
       "      <th>area_mean</th>\n",
       "      <th>smoothness_mean</th>\n",
       "      <th>compactness_mean</th>\n",
       "      <th>concavity_mean</th>\n",
       "      <th>concave_pts_mean</th>\n",
       "      <th>symmetry_mean</th>\n",
       "      <th>...</th>\n",
       "      <th>texture_worst</th>\n",
       "      <th>perimeter_worst</th>\n",
       "      <th>area_worst</th>\n",
       "      <th>smoothness_worst</th>\n",
       "      <th>compactness_worst</th>\n",
       "      <th>concavity_worst</th>\n",
       "      <th>concave_pts_worst</th>\n",
       "      <th>symmetry_worst</th>\n",
       "      <th>fractal_dim_worst</th>\n",
       "      <th>diagnostic</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>561</td>\n",
       "      <td>20.47</td>\n",
       "      <td>20.67</td>\n",
       "      <td>134.70</td>\n",
       "      <td>1299.0</td>\n",
       "      <td>0.09156</td>\n",
       "      <td>0.13130</td>\n",
       "      <td>0.15230</td>\n",
       "      <td>0.10150</td>\n",
       "      <td>0.2166</td>\n",
       "      <td>...</td>\n",
       "      <td>27.15</td>\n",
       "      <td>152.00</td>\n",
       "      <td>1645.0</td>\n",
       "      <td>0.1097</td>\n",
       "      <td>0.25340</td>\n",
       "      <td>0.30920</td>\n",
       "      <td>0.16130</td>\n",
       "      <td>0.3220</td>\n",
       "      <td>0.06386</td>\n",
       "      <td>M</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>77</td>\n",
       "      <td>11.75</td>\n",
       "      <td>20.18</td>\n",
       "      <td>76.10</td>\n",
       "      <td>419.8</td>\n",
       "      <td>0.10890</td>\n",
       "      <td>0.11410</td>\n",
       "      <td>0.06843</td>\n",
       "      <td>0.03738</td>\n",
       "      <td>0.1993</td>\n",
       "      <td>...</td>\n",
       "      <td>26.21</td>\n",
       "      <td>88.91</td>\n",
       "      <td>543.9</td>\n",
       "      <td>0.1358</td>\n",
       "      <td>0.18920</td>\n",
       "      <td>0.19560</td>\n",
       "      <td>0.07909</td>\n",
       "      <td>0.3168</td>\n",
       "      <td>0.07987</td>\n",
       "      <td>B</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>151</td>\n",
       "      <td>13.77</td>\n",
       "      <td>13.27</td>\n",
       "      <td>88.06</td>\n",
       "      <td>582.7</td>\n",
       "      <td>0.09198</td>\n",
       "      <td>0.06221</td>\n",
       "      <td>0.01063</td>\n",
       "      <td>0.01917</td>\n",
       "      <td>0.1592</td>\n",
       "      <td>...</td>\n",
       "      <td>16.93</td>\n",
       "      <td>94.17</td>\n",
       "      <td>661.1</td>\n",
       "      <td>0.1170</td>\n",
       "      <td>0.10720</td>\n",
       "      <td>0.03732</td>\n",
       "      <td>0.05802</td>\n",
       "      <td>0.2823</td>\n",
       "      <td>0.06794</td>\n",
       "      <td>B</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>231</td>\n",
       "      <td>12.34</td>\n",
       "      <td>14.95</td>\n",
       "      <td>78.29</td>\n",
       "      <td>469.1</td>\n",
       "      <td>0.08682</td>\n",
       "      <td>0.04571</td>\n",
       "      <td>0.02109</td>\n",
       "      <td>0.02054</td>\n",
       "      <td>0.1571</td>\n",
       "      <td>...</td>\n",
       "      <td>16.85</td>\n",
       "      <td>84.11</td>\n",
       "      <td>533.1</td>\n",
       "      <td>0.1048</td>\n",
       "      <td>0.06744</td>\n",
       "      <td>0.04921</td>\n",
       "      <td>0.04793</td>\n",
       "      <td>0.2298</td>\n",
       "      <td>0.05974</td>\n",
       "      <td>B</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>410</td>\n",
       "      <td>13.80</td>\n",
       "      <td>15.79</td>\n",
       "      <td>90.43</td>\n",
       "      <td>584.1</td>\n",
       "      <td>0.10070</td>\n",
       "      <td>0.12800</td>\n",
       "      <td>0.07789</td>\n",
       "      <td>0.05069</td>\n",
       "      <td>0.1662</td>\n",
       "      <td>...</td>\n",
       "      <td>20.86</td>\n",
       "      <td>110.30</td>\n",
       "      <td>812.4</td>\n",
       "      <td>0.1411</td>\n",
       "      <td>0.35420</td>\n",
       "      <td>0.27790</td>\n",
       "      <td>0.13830</td>\n",
       "      <td>0.2589</td>\n",
       "      <td>0.10300</td>\n",
       "      <td>M</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 32 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    id  radius_mean  texture_mean  perimeter_mean  area_mean  smoothness_mean  \\\n",
       "0  561        20.47         20.67          134.70     1299.0          0.09156   \n",
       "1   77        11.75         20.18           76.10      419.8          0.10890   \n",
       "2  151        13.77         13.27           88.06      582.7          0.09198   \n",
       "3  231        12.34         14.95           78.29      469.1          0.08682   \n",
       "4  410        13.80         15.79           90.43      584.1          0.10070   \n",
       "\n",
       "   compactness_mean  concavity_mean  concave_pts_mean  symmetry_mean  ...  \\\n",
       "0           0.13130         0.15230           0.10150         0.2166  ...   \n",
       "1           0.11410         0.06843           0.03738         0.1993  ...   \n",
       "2           0.06221         0.01063           0.01917         0.1592  ...   \n",
       "3           0.04571         0.02109           0.02054         0.1571  ...   \n",
       "4           0.12800         0.07789           0.05069         0.1662  ...   \n",
       "\n",
       "   texture_worst  perimeter_worst  area_worst  smoothness_worst  \\\n",
       "0          27.15           152.00      1645.0            0.1097   \n",
       "1          26.21            88.91       543.9            0.1358   \n",
       "2          16.93            94.17       661.1            0.1170   \n",
       "3          16.85            84.11       533.1            0.1048   \n",
       "4          20.86           110.30       812.4            0.1411   \n",
       "\n",
       "   compactness_worst  concavity_worst  concave_pts_worst  symmetry_worst  \\\n",
       "0            0.25340          0.30920            0.16130          0.3220   \n",
       "1            0.18920          0.19560            0.07909          0.3168   \n",
       "2            0.10720          0.03732            0.05802          0.2823   \n",
       "3            0.06744          0.04921            0.04793          0.2298   \n",
       "4            0.35420          0.27790            0.13830          0.2589   \n",
       "\n",
       "   fractal_dim_worst  diagnostic  \n",
       "0            0.06386           M  \n",
       "1            0.07987           B  \n",
       "2            0.06794           B  \n",
       "3            0.05974           B  \n",
       "4            0.10300           M  \n",
       "\n",
       "[5 rows x 32 columns]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df1 = pd.read_csv('data/dataset1.csv')\n",
    "print(\"Shape:\", df1.shape)\n",
    "df1.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2e43cf3e",
   "metadata": {},
   "source": [
    "### Dataset 2\n",
    "\n",
    "[source](https://www.kaggle.com/code/a3amat02/breast-cancer-classification/input)\n",
    "\n",
    "**About this file**\n",
    "\n",
    "* y. The outcomes. A factor with two levels denoting whether a mass is malignant (\"M\") or benign (\"B\").\n",
    "* x. The predictors. A matrix with the mean, standard error and worst value of each of 10 nuclear measurements on the slide, for 30 total features per biopsy:\n",
    "* radius. Nucleus radius (mean of distances from center to points on perimeter).\n",
    "* texture. Nucleus texture (standard deviation of grayscale values).\n",
    "* perimeter. Nucleus perimeter.\n",
    "* area. Nucleus area.\n",
    "* smoothness. Nucleus smoothness (local variation in radius lengths).\n",
    "* compactness. Nucleus compactness (perimeter^2/area - 1).\n",
    "* concavity, Nucleus concavity (severity of concave portions of the contour).\n",
    "* concave_pts. Number of concave portions of the nucleus contour.\n",
    "* symmetry. Nucleus symmetry.\n",
    "* fractal_dim. Nucleus fractal dimension (\"coastline approximation\" -1)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "ead81772",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape: (1193, 32)\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>id</th>\n",
       "      <th>radius_mean</th>\n",
       "      <th>texture_mean</th>\n",
       "      <th>perimeter_mean</th>\n",
       "      <th>area_mean</th>\n",
       "      <th>smoothness_mean</th>\n",
       "      <th>compactness_mean</th>\n",
       "      <th>concavity_mean</th>\n",
       "      <th>concave_pts_mean</th>\n",
       "      <th>symmetry_mean</th>\n",
       "      <th>...</th>\n",
       "      <th>texture_worst</th>\n",
       "      <th>perimeter_worst</th>\n",
       "      <th>area_worst</th>\n",
       "      <th>smoothness_worst</th>\n",
       "      <th>compactness_worst</th>\n",
       "      <th>concavity_worst</th>\n",
       "      <th>concave_pts_worst</th>\n",
       "      <th>symmetry_worst</th>\n",
       "      <th>fractal_dim_worst</th>\n",
       "      <th>diagnostic</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>374</td>\n",
       "      <td>14.68</td>\n",
       "      <td>20.13</td>\n",
       "      <td>94.74</td>\n",
       "      <td>684.5</td>\n",
       "      <td>0.09867</td>\n",
       "      <td>0.07200</td>\n",
       "      <td>0.073950</td>\n",
       "      <td>0.052590</td>\n",
       "      <td>0.1586</td>\n",
       "      <td>...</td>\n",
       "      <td>30.88</td>\n",
       "      <td>123.40</td>\n",
       "      <td>1138.0</td>\n",
       "      <td>0.14640</td>\n",
       "      <td>0.18710</td>\n",
       "      <td>0.29140</td>\n",
       "      <td>0.16090</td>\n",
       "      <td>0.3029</td>\n",
       "      <td>0.08216</td>\n",
       "      <td>M</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>325</td>\n",
       "      <td>12.34</td>\n",
       "      <td>12.27</td>\n",
       "      <td>78.94</td>\n",
       "      <td>468.5</td>\n",
       "      <td>0.09003</td>\n",
       "      <td>0.06307</td>\n",
       "      <td>0.029580</td>\n",
       "      <td>0.026470</td>\n",
       "      <td>0.1689</td>\n",
       "      <td>...</td>\n",
       "      <td>19.27</td>\n",
       "      <td>87.22</td>\n",
       "      <td>564.9</td>\n",
       "      <td>0.12920</td>\n",
       "      <td>0.20740</td>\n",
       "      <td>0.17910</td>\n",
       "      <td>0.10700</td>\n",
       "      <td>0.3110</td>\n",
       "      <td>0.07592</td>\n",
       "      <td>B</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>359</td>\n",
       "      <td>20.57</td>\n",
       "      <td>17.77</td>\n",
       "      <td>132.90</td>\n",
       "      <td>1326.0</td>\n",
       "      <td>0.08474</td>\n",
       "      <td>0.07864</td>\n",
       "      <td>0.086900</td>\n",
       "      <td>0.070170</td>\n",
       "      <td>0.1812</td>\n",
       "      <td>...</td>\n",
       "      <td>23.41</td>\n",
       "      <td>158.80</td>\n",
       "      <td>1956.0</td>\n",
       "      <td>0.12380</td>\n",
       "      <td>0.18660</td>\n",
       "      <td>0.24160</td>\n",
       "      <td>0.18600</td>\n",
       "      <td>0.2750</td>\n",
       "      <td>0.08902</td>\n",
       "      <td>M</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>264</td>\n",
       "      <td>14.80</td>\n",
       "      <td>17.66</td>\n",
       "      <td>95.88</td>\n",
       "      <td>674.8</td>\n",
       "      <td>0.09179</td>\n",
       "      <td>0.08890</td>\n",
       "      <td>0.040690</td>\n",
       "      <td>0.022600</td>\n",
       "      <td>0.1893</td>\n",
       "      <td>...</td>\n",
       "      <td>22.74</td>\n",
       "      <td>105.90</td>\n",
       "      <td>829.5</td>\n",
       "      <td>0.12260</td>\n",
       "      <td>0.18810</td>\n",
       "      <td>0.20600</td>\n",
       "      <td>0.08308</td>\n",
       "      <td>0.3600</td>\n",
       "      <td>0.07285</td>\n",
       "      <td>B</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>161</td>\n",
       "      <td>13.50</td>\n",
       "      <td>12.71</td>\n",
       "      <td>85.69</td>\n",
       "      <td>566.2</td>\n",
       "      <td>0.07376</td>\n",
       "      <td>0.03614</td>\n",
       "      <td>0.002758</td>\n",
       "      <td>0.004419</td>\n",
       "      <td>0.1365</td>\n",
       "      <td>...</td>\n",
       "      <td>16.94</td>\n",
       "      <td>95.48</td>\n",
       "      <td>698.7</td>\n",
       "      <td>0.09023</td>\n",
       "      <td>0.05836</td>\n",
       "      <td>0.01379</td>\n",
       "      <td>0.02210</td>\n",
       "      <td>0.2267</td>\n",
       "      <td>0.06192</td>\n",
       "      <td>B</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 32 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    id  radius_mean  texture_mean  perimeter_mean  area_mean  smoothness_mean  \\\n",
       "0  374        14.68         20.13           94.74      684.5          0.09867   \n",
       "1  325        12.34         12.27           78.94      468.5          0.09003   \n",
       "2  359        20.57         17.77          132.90     1326.0          0.08474   \n",
       "3  264        14.80         17.66           95.88      674.8          0.09179   \n",
       "4  161        13.50         12.71           85.69      566.2          0.07376   \n",
       "\n",
       "   compactness_mean  concavity_mean  concave_pts_mean  symmetry_mean  ...  \\\n",
       "0           0.07200        0.073950          0.052590         0.1586  ...   \n",
       "1           0.06307        0.029580          0.026470         0.1689  ...   \n",
       "2           0.07864        0.086900          0.070170         0.1812  ...   \n",
       "3           0.08890        0.040690          0.022600         0.1893  ...   \n",
       "4           0.03614        0.002758          0.004419         0.1365  ...   \n",
       "\n",
       "   texture_worst  perimeter_worst  area_worst  smoothness_worst  \\\n",
       "0          30.88           123.40      1138.0           0.14640   \n",
       "1          19.27            87.22       564.9           0.12920   \n",
       "2          23.41           158.80      1956.0           0.12380   \n",
       "3          22.74           105.90       829.5           0.12260   \n",
       "4          16.94            95.48       698.7           0.09023   \n",
       "\n",
       "   compactness_worst  concavity_worst  concave_pts_worst  symmetry_worst  \\\n",
       "0            0.18710          0.29140            0.16090          0.3029   \n",
       "1            0.20740          0.17910            0.10700          0.3110   \n",
       "2            0.18660          0.24160            0.18600          0.2750   \n",
       "3            0.18810          0.20600            0.08308          0.3600   \n",
       "4            0.05836          0.01379            0.02210          0.2267   \n",
       "\n",
       "   fractal_dim_worst  diagnostic  \n",
       "0            0.08216           M  \n",
       "1            0.07592           B  \n",
       "2            0.08902           M  \n",
       "3            0.07285           B  \n",
       "4            0.06192           B  \n",
       "\n",
       "[5 rows x 32 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df2 = pd.read_csv('data/dataset2.csv')\n",
    "print(\"Shape:\", df2.shape)\n",
    "df2.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a0ac0ad7",
   "metadata": {},
   "source": [
    "### Dataset 3\n",
    "\n",
    "[source](https://archive.ics.uci.edu/ml/datasets/Breast+Cancer+Wisconsin+(Diagnostic))\n",
    "\n",
    "**Data Set Information:**\n",
    "\n",
    "Features are computed from a digitized image of a fine needle aspirate (FNA) of a breast mass. They describe characteristics of the cell nuclei present in the image. A few of the images can be found at [Web Link]\n",
    "\n",
    "Separating plane described above was obtained using Multisurface Method-Tree (MSM-T) [K. P. Bennett, \"Decision Tree Construction Via Linear Programming.\" Proceedings of the 4th Midwest Artificial Intelligence and Cognitive Science Society, pp. 97-101, 1992], a classification method which uses linear programming to construct a decision tree. Relevant features were selected using an exhaustive search in the space of 1-4 features and 1-3 separating planes.\n",
    "\n",
    "The actual linear program used to obtain the separating plane in the 3-dimensional space is that described in: [K. P. Bennett and O. L. Mangasarian: \"Robust Linear Programming Discrimination of Two Linearly Inseparable Sets\", Optimization Methods and Software 1, 1992, 23-34].\n",
    "\n",
    "This database is also available through the UW CS ftp server:\n",
    "ftp ftp.cs.wisc.edu\n",
    "cd math-prog/cpo-dataset/machine-learn/WDBC/\n",
    "\n",
    "\n",
    "**Attribute Information:**\n",
    "\n",
    "1) ID number\n",
    "\n",
    "2) Diagnosis (M = malignant, B = benign)\n",
    "\n",
    "3-32)\n",
    "\n",
    "Ten real-valued features are computed for each cell nucleus:\n",
    "\n",
    "a) radius (mean of distances from center to points on the perimeter)\n",
    "\n",
    "b) texture (standard deviation of gray-scale values)\n",
    "\n",
    "c) perimeter\n",
    "\n",
    "d) area\n",
    "\n",
    "e) smoothness (local variation in radius lengths)\n",
    "\n",
    "f) compactness (perimeter^2 / area - 1.0)\n",
    "\n",
    "g) concavity (severity of concave portions of the contour)\n",
    "\n",
    "h) concave points (number of concave portions of the contour)\n",
    "\n",
    "i) symmetry\n",
    "\n",
    "j) fractal dimension (\"coastline approximation\" - 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "d995bdb5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape: (1189, 32)\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>id</th>\n",
       "      <th>radius_mean</th>\n",
       "      <th>texture_mean</th>\n",
       "      <th>perimeter_mean</th>\n",
       "      <th>area_mean</th>\n",
       "      <th>smoothness_mean</th>\n",
       "      <th>compactness_mean</th>\n",
       "      <th>concavity_mean</th>\n",
       "      <th>concave_pts_mean</th>\n",
       "      <th>symmetry_mean</th>\n",
       "      <th>...</th>\n",
       "      <th>texture_worst</th>\n",
       "      <th>perimeter_worst</th>\n",
       "      <th>area_worst</th>\n",
       "      <th>smoothness_worst</th>\n",
       "      <th>compactness_worst</th>\n",
       "      <th>concavity_worst</th>\n",
       "      <th>concave_pts_worst</th>\n",
       "      <th>symmetry_worst</th>\n",
       "      <th>fractal_dim_worst</th>\n",
       "      <th>diagnostic</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>529</td>\n",
       "      <td>15.49</td>\n",
       "      <td>19.97</td>\n",
       "      <td>102.40</td>\n",
       "      <td>744.7</td>\n",
       "      <td>0.11600</td>\n",
       "      <td>0.15620</td>\n",
       "      <td>0.18910</td>\n",
       "      <td>0.09113</td>\n",
       "      <td>0.1929</td>\n",
       "      <td>...</td>\n",
       "      <td>29.41</td>\n",
       "      <td>142.10</td>\n",
       "      <td>1359.0</td>\n",
       "      <td>0.1681</td>\n",
       "      <td>0.3913</td>\n",
       "      <td>0.55530</td>\n",
       "      <td>0.21210</td>\n",
       "      <td>0.3187</td>\n",
       "      <td>0.10190</td>\n",
       "      <td>M</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>365</td>\n",
       "      <td>13.71</td>\n",
       "      <td>20.83</td>\n",
       "      <td>90.20</td>\n",
       "      <td>577.9</td>\n",
       "      <td>0.11890</td>\n",
       "      <td>0.16450</td>\n",
       "      <td>0.09366</td>\n",
       "      <td>0.05985</td>\n",
       "      <td>0.2196</td>\n",
       "      <td>...</td>\n",
       "      <td>28.14</td>\n",
       "      <td>110.60</td>\n",
       "      <td>897.0</td>\n",
       "      <td>0.1654</td>\n",
       "      <td>0.3682</td>\n",
       "      <td>0.26780</td>\n",
       "      <td>0.15560</td>\n",
       "      <td>0.3196</td>\n",
       "      <td>0.11510</td>\n",
       "      <td>M</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>447</td>\n",
       "      <td>13.43</td>\n",
       "      <td>19.63</td>\n",
       "      <td>85.84</td>\n",
       "      <td>565.4</td>\n",
       "      <td>0.09048</td>\n",
       "      <td>0.06288</td>\n",
       "      <td>0.05858</td>\n",
       "      <td>0.03438</td>\n",
       "      <td>0.1598</td>\n",
       "      <td>...</td>\n",
       "      <td>29.87</td>\n",
       "      <td>116.60</td>\n",
       "      <td>993.6</td>\n",
       "      <td>0.1401</td>\n",
       "      <td>0.1546</td>\n",
       "      <td>0.26440</td>\n",
       "      <td>0.11600</td>\n",
       "      <td>0.2884</td>\n",
       "      <td>0.07371</td>\n",
       "      <td>M</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>347</td>\n",
       "      <td>12.77</td>\n",
       "      <td>29.43</td>\n",
       "      <td>81.35</td>\n",
       "      <td>507.9</td>\n",
       "      <td>0.08276</td>\n",
       "      <td>0.04234</td>\n",
       "      <td>0.01997</td>\n",
       "      <td>0.01499</td>\n",
       "      <td>0.1539</td>\n",
       "      <td>...</td>\n",
       "      <td>36.00</td>\n",
       "      <td>88.10</td>\n",
       "      <td>594.7</td>\n",
       "      <td>0.1234</td>\n",
       "      <td>0.1064</td>\n",
       "      <td>0.08653</td>\n",
       "      <td>0.06498</td>\n",
       "      <td>0.2407</td>\n",
       "      <td>0.06484</td>\n",
       "      <td>B</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>27</td>\n",
       "      <td>12.00</td>\n",
       "      <td>15.65</td>\n",
       "      <td>76.95</td>\n",
       "      <td>443.3</td>\n",
       "      <td>0.09723</td>\n",
       "      <td>0.07165</td>\n",
       "      <td>0.04151</td>\n",
       "      <td>0.01863</td>\n",
       "      <td>0.2079</td>\n",
       "      <td>...</td>\n",
       "      <td>24.90</td>\n",
       "      <td>87.78</td>\n",
       "      <td>567.9</td>\n",
       "      <td>0.1377</td>\n",
       "      <td>0.2003</td>\n",
       "      <td>0.22670</td>\n",
       "      <td>0.07632</td>\n",
       "      <td>0.3379</td>\n",
       "      <td>0.07924</td>\n",
       "      <td>B</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 32 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    id  radius_mean  texture_mean  perimeter_mean  area_mean  smoothness_mean  \\\n",
       "0  529        15.49         19.97          102.40      744.7          0.11600   \n",
       "1  365        13.71         20.83           90.20      577.9          0.11890   \n",
       "2  447        13.43         19.63           85.84      565.4          0.09048   \n",
       "3  347        12.77         29.43           81.35      507.9          0.08276   \n",
       "4   27        12.00         15.65           76.95      443.3          0.09723   \n",
       "\n",
       "   compactness_mean  concavity_mean  concave_pts_mean  symmetry_mean  ...  \\\n",
       "0           0.15620         0.18910           0.09113         0.1929  ...   \n",
       "1           0.16450         0.09366           0.05985         0.2196  ...   \n",
       "2           0.06288         0.05858           0.03438         0.1598  ...   \n",
       "3           0.04234         0.01997           0.01499         0.1539  ...   \n",
       "4           0.07165         0.04151           0.01863         0.2079  ...   \n",
       "\n",
       "   texture_worst  perimeter_worst  area_worst  smoothness_worst  \\\n",
       "0          29.41           142.10      1359.0            0.1681   \n",
       "1          28.14           110.60       897.0            0.1654   \n",
       "2          29.87           116.60       993.6            0.1401   \n",
       "3          36.00            88.10       594.7            0.1234   \n",
       "4          24.90            87.78       567.9            0.1377   \n",
       "\n",
       "   compactness_worst  concavity_worst  concave_pts_worst  symmetry_worst  \\\n",
       "0             0.3913          0.55530            0.21210          0.3187   \n",
       "1             0.3682          0.26780            0.15560          0.3196   \n",
       "2             0.1546          0.26440            0.11600          0.2884   \n",
       "3             0.1064          0.08653            0.06498          0.2407   \n",
       "4             0.2003          0.22670            0.07632          0.3379   \n",
       "\n",
       "   fractal_dim_worst  diagnostic  \n",
       "0            0.10190           M  \n",
       "1            0.11510           M  \n",
       "2            0.07371           M  \n",
       "3            0.06484           B  \n",
       "4            0.07924           B  \n",
       "\n",
       "[5 rows x 32 columns]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df3 = pd.read_csv('data/dataset3.csv')\n",
    "print(\"Shape:\", df3.shape)\n",
    "df3.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "45cc3eaf",
   "metadata": {},
   "source": [
    "# Data preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "df47c07e",
   "metadata": {},
   "outputs": [],
   "source": [
    "def scale_dataset(df, overSample=False):\n",
    "    # split to fetures and diagnostic result\n",
    "    X = df[df.columns[:-1]].values\n",
    "    Y = df[df.columns[-1]].values\n",
    "    \n",
    "    # standardize the input features \n",
    "    scaler = StandardScaler()\n",
    "    X = scaler.fit_transform(X)\n",
    "    \n",
    "    # balance the class distribution\n",
    "    if overSample:\n",
    "        ros = RandomOverSampler()\n",
    "        X, Y = ros.fit_resample(X, Y)\n",
    "\n",
    "    data = np.hstack((X, np.reshape(Y, (-1, 1))))\n",
    "    \n",
    "    # convert to tensor context\n",
    "    X_train_tensor = Variable(torch.tensor(X, dtype = torch.float32))\n",
    "    Y_train_tensor = Variable(torch.tensor(Y, dtype = torch.float32))\n",
    "    data_tensor    = Variable(torch.tensor(data, dtype = torch.float32))\n",
    "    \n",
    "    return data_tensor, X_train_tensor, Y_train_tensor"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "25366206",
   "metadata": {},
   "source": [
    "# Logistic Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "6bf48c94",
   "metadata": {},
   "outputs": [],
   "source": [
    "class LogisticRegression(torch.nn.Module):\n",
    "    def __init__(self, num_features):\n",
    "        # init super class of LogisticRegression\n",
    "        super(LogisticRegression, self).__init__()\n",
    "        \n",
    "        # create \"linear neural network\"\n",
    "        input_dim  = num_features\n",
    "        output_dim = 1\n",
    "        self.linear = torch.nn.Linear(input_dim, output_dim)\n",
    "\n",
    "        # initialize Weights and Bias\n",
    "        self.linear.weight.detach().zero_()\n",
    "        self.linear.bias.detach().zero_()\n",
    "\n",
    "    def forward(self, x):\n",
    "        return torch.sigmoid(self.linear(x))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f38de85a",
   "metadata": {},
   "source": [
    "In this demo, there are 3 hospitals (with corresponding datasets above) but there could be more hospitals. The 3 hospitals cannot share the cases of their patients because they are competitors and it is necessary to protect the privacy of patients. Hence, the ML model will be learned in a federated way."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "71fa9a72",
   "metadata": {},
   "source": [
    "# Client"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "38956407",
   "metadata": {},
   "source": [
    "Import openFHE library. Then run `generate_key()`, this is just a python wrapper to run `openfhe-lib/build/key_gen`. After running this files, it will create `openfhe-lib/data` folder that holds:\n",
    "- `crypto_context.txt` : contain BGV CryptoContext object\n",
    "- `public_key.txt`     : contain BGV public key\n",
    "- `private_key.txt`    : contain BGV private key\n",
    "- `mult_key.txt`       : contain BGV multiplication key"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "727118a0",
   "metadata": {},
   "outputs": [],
   "source": [
    "from openfhe_lib.bfv.openFHE import *\n",
    "\n",
    "# === Generate Key-pairs of BGV Context ===\n",
    "generate_keys()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a0715f90",
   "metadata": {},
   "source": [
    "First, we start by creating the `Client` class that simulate the computers of each hospital. This just rewrite the Logistic Regression process that we have implemented previously."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "87fa6dae",
   "metadata": {},
   "outputs": [],
   "source": [
    "class Client:\n",
    "    def __init__(self, name, data_url, enc_file, n_features, iters):\n",
    "        self.id = name\n",
    "        self.enc_file = enc_file  # place wher clients save encrypted weights\n",
    "        \n",
    "        # split data into train and test\n",
    "        self.X_train, self.Y_train, self.X_test, self.Y_test = self.preprocessing(data_url)\n",
    "        \n",
    "        # define local training model\n",
    "        self.local_model = LogisticRegression(n_features)\n",
    "        \n",
    "        # some helpfull stuffs\n",
    "        self.decide_vectorized = np.vectorize(self.decide)\n",
    "        self.to_percent = lambda x: '{:.2f}%'.format(x)\n",
    "        self.num_epochs = iters\n",
    "        self.accuracies = []\n",
    "        self.losses = []\n",
    "        \n",
    "    def preprocessing(self, data_url):\n",
    "        df = pd.read_csv(data_url)\n",
    "        # Replace \"M\" with 1 and \"B\" with 0 at \"diagnostic\" column\n",
    "        df[\"diagnostic\"] = (df[\"diagnostic\"] == \"M\").astype(int)\n",
    "        \n",
    "        # split dataframe to train and test df\n",
    "        df_train, df_test = np.split(df.sample(frac=1), [int(0.8 * len(df))])\n",
    "        \n",
    "        # scaling and convert to tensor context\n",
    "        train, X_train, Y_train = scale_dataset(df_train, True)\n",
    "        test , X_test , Y_test  = scale_dataset(df_test , False)\n",
    "        return X_train, Y_train, X_test, Y_test\n",
    "    \n",
    "    def decide(self, y):\n",
    "        return 1. if y >= 0.5 else 0.\n",
    "    \n",
    "    def compute_accuracy(self, input, output):\n",
    "        prediction = self.local_model(input).data.numpy()[:, 0]\n",
    "        n_samples = prediction.shape[0] + 0.\n",
    "        prediction = self.decide_vectorized(prediction)\n",
    "        equal = prediction == output.data.numpy()\n",
    "        return 100. * equal.sum() / n_samples\n",
    "    \n",
    "    def local_training(self, debug=True):\n",
    "        n_samples, _ = self.X_train.shape\n",
    "\n",
    "        # define criterion function and set up optimizer\n",
    "        criterion = torch.nn.BCELoss(reduction='mean')\n",
    "        optimizer = torch.optim.SGD(self.local_model.parameters(), lr=0.01)  \n",
    "\n",
    "        # main process\n",
    "        for epoch in range(self.num_epochs):  \n",
    "            optimizer.zero_grad()\n",
    "            #### Compute outputs ####\n",
    "            prediction = self.local_model(self.X_train)\n",
    "\n",
    "            #### Compute gradients ####\n",
    "            loss = criterion(prediction.squeeze(), self.Y_train)\n",
    "            loss.backward()\n",
    "\n",
    "            #### Update weights #### \n",
    "            optimizer.step()\n",
    "\n",
    "            # compute accuracy and loss\n",
    "            train_acc = self.compute_accuracy(self.X_train, self.Y_train)\n",
    "            train_loss = loss.item()\n",
    "                \n",
    "            self.losses.append(train_loss)\n",
    "            self.accuracies.append(train_acc)\n",
    "        \n",
    "            #### Logging ####\n",
    "            if debug and (epoch + 1)%50 == 0:\n",
    "                print('[LOG] Epoch: %05d' % (epoch + 1), end=\"\")\n",
    "                print('    | Train ACC: %s' % self.to_percent(train_acc), end=\"\")\n",
    "                print('    | Loss: %.3f' % train_loss)\n",
    "    \n",
    "    def encrypted_model_params(self):\n",
    "        model_weights = self.local_model.linear.weight.data.squeeze().tolist()\n",
    "        model_bias    = self.local_model.linear.bias.data.squeeze().tolist()\n",
    "        \n",
    "        model_params  = model_weights + [model_bias]\n",
    "        encrypt_weights(model_params, self.enc_file)\n",
    "        \n",
    "    def decrypted_model_params(self):\n",
    "        params = decrypt_weights(\"/enc_aggregator_weight_server.txt\")\n",
    "        # convert float to tensor context\n",
    "        W = Variable(torch.tensor([params[:-1]], dtype = torch.float32))\n",
    "        B = Variable(torch.tensor( params[-1], dtype = torch.float32))\n",
    "        \n",
    "        self.local_model.linear.weight = torch.nn.Parameter(W)\n",
    "        self.local_model.linear.bias   = torch.nn.Parameter(B)\n",
    "    \n",
    "    def plot_graphs(self, diagnosis_title = 'Breast cancer'):\n",
    "        plt.plot(self.losses)\n",
    "        plt.title(f\"{diagnosis_title} - Training Loss\")\n",
    "        plt.xlabel(\"Iterations\")\n",
    "        plt.ylabel(\"Training Loss\")\n",
    "        plt.show()\n",
    "        plt.plot(self.accuracies)\n",
    "        plt.title(f\"{diagnosis_title} - Training Accuracy\")\n",
    "        plt.xlabel(\"Iterations\")\n",
    "        plt.ylabel(\"Training Accuracy (Percent %)\")\n",
    "        plt.show()\n",
    "    \n",
    "    def print_result_after_training(self):\n",
    "        print('Model parameters:')\n",
    "        print('  | Weights: %s' % self.local_model.linear.weight)\n",
    "        print('  | Bias: %s' % self.local_model.linear.bias)\n",
    "        self.plot_graphs()\n",
    "    \n",
    "    def evaluating_model(self):\n",
    "        test_acc = self.compute_accuracy(self.X_test, self.Y_test)\n",
    "        print('[+] Testing Accuracy = {}'.format(self.to_percent(test_acc)))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "73f26b13",
   "metadata": {},
   "source": [
    "# Server"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a2ba0328",
   "metadata": {},
   "source": [
    "We define some functions to train the machine learning model in a federated way while keeping track of the training loss and the training accuracy, for each hospital separately."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "3f52071e",
   "metadata": {},
   "outputs": [],
   "source": [
    "clients = [\n",
    "    Client('Hostpital1', 'data/dataset1.csv', \"/enc_weight_client1.txt\", n_features=31, iters=10), \n",
    "    Client('Hostpital2', 'data/dataset2.csv', \"/enc_weight_client2.txt\", n_features=31, iters=10),\n",
    "    Client('Hostpital3', 'data/dataset3.csv', \"/enc_weight_client3.txt\", n_features=31, iters=10),\n",
    "    Client('Hostpital4', 'data/dataset4.csv', \"/enc_weight_client4.txt\", n_features=31, iters=10)\n",
    "]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f76d8320",
   "metadata": {},
   "source": [
    "The whole process is done in a server aggregator, in 1000 iterations (we can vary the number of iterations.) At each iteration."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "f2fc3c60",
   "metadata": {},
   "outputs": [],
   "source": [
    "from tqdm.notebook import tqdm\n",
    "\n",
    "iterations = 1000 #2000\n",
    "worker_iterations = 5\n",
    "to_percent = lambda x: '{:.2f}%'.format(x)\n",
    "n_hospitals = len(clients)\n",
    "n_features = 31\n",
    "    \n",
    "def compute_federated_accuracy(model, input, output):\n",
    "    prediction = model(input)\n",
    "    n_samples = prediction.shape[0]\n",
    "    s = 0.\n",
    "    for i in range(n_samples):\n",
    "        p = 1. if prediction[i] >= 0.5 else 0.\n",
    "        e = 1. if p == output[i] else 0.\n",
    "        s += e\n",
    "    return 100. * s / n_samples\n",
    "\n",
    "def federated_learning(clients):\n",
    "    # init global training model\n",
    "    global_model = LogisticRegression(n_features)\n",
    "\n",
    "    # record losses and accuracies report from clients\n",
    "    losses = [[] for i in range(n_hospitals)]\n",
    "    accuracies = [[] for i in range(n_hospitals)]\n",
    "    \n",
    "    pbar = tqdm(range(iterations), desc='Federated Learning Process')\n",
    "    for iteration in pbar:\n",
    "        if iteration:\n",
    "            # copy global model to clients\n",
    "            # clients will receive the weight-aggregated from server, extract\n",
    "            # the ciphertext then decrypt it to get to global_model's weights\n",
    "            for i in range(n_hospitals):\n",
    "                clients[i].decrypted_model_params()\n",
    "        \n",
    "        # perform local training for each clients then report acc and loss to server\n",
    "        for i in range(n_hospitals):\n",
    "            clients[i].local_training(debug=False)\n",
    "            \n",
    "            # report to server\n",
    "            losses[i].append(clients[i].losses[-1])\n",
    "            accuracies[i].append(clients[i].accuracies[-1])\n",
    "        \n",
    "        # clients encrypt the final weights of local model after training\n",
    "        for i in range(n_hospitals):\n",
    "            clients[i].encrypted_model_params()\n",
    "        \n",
    "        # server collect clients's encrypted weights then perform weight-aggregation\n",
    "        # by using homomorphic operation\n",
    "        with torch.no_grad():\n",
    "            # avg_weight = sum([clients[i].local_model.linear.weight.data for i in range(n_hospitals)]) / n_hospitals\n",
    "            # global_model.linear.weight = nn.Parameter(avg_weight)\n",
    "            # avg_bias = sum([clients[i].local_model.linear.bias.data for i in range(n_hospitals)]) / n_hospitals\n",
    "            # global_model.linear.bias = nn.Parameter(avg_bias)\n",
    "            aggregator()\n",
    "    \n",
    "        # logging\n",
    "        if (iteration + 1) % 100 == 0:\n",
    "            losses_str = ['{:.4f}'.format(losses[i][-1]) for i in range(n_hospitals)]\n",
    "            accuracies_str = [to_percent(accuracies[i][-1]) for i in range(n_hospitals)]\n",
    "            print('[LOG] Epoch = {0:04d}\\n> Losses = {1}\\n> Accuracies = {2}'.format(iteration + 1, losses_str, accuracies_str))\n",
    "        \n",
    "    return losses, accuracies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "8f859db4",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7a5a23aba83f4a0c942a31d317b0fc30",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Federated Learning Process:   0%|          | 0/1000 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[LOG] Epoch = 0100\n",
      "> Losses = ['0.0759', '0.0576', '0.0664', '0.0838']\n",
      "> Accuracies = ['97.72%', '98.65%', '98.34%', '97.22%']\n",
      "[LOG] Epoch = 0200\n",
      "> Losses = ['0.0600', '0.0413', '0.0505', '0.0657']\n",
      "> Accuracies = ['98.02%', '98.65%', '98.76%', '97.50%']\n",
      "[LOG] Epoch = 0300\n",
      "> Losses = ['0.0529', '0.0344', '0.0436', '0.0571']\n",
      "> Accuracies = ['98.61%', '98.86%', '99.17%', '98.06%']\n",
      "[LOG] Epoch = 0400\n",
      "> Losses = ['0.0488', '0.0305', '0.0396', '0.0518']\n",
      "> Accuracies = ['98.61%', '99.07%', '99.17%', '98.06%']\n",
      "[LOG] Epoch = 0500\n",
      "> Losses = ['0.0460', '0.0279', '0.0368', '0.0482']\n",
      "> Accuracies = ['98.81%', '99.28%', '99.17%', '98.33%']\n",
      "[LOG] Epoch = 0600\n",
      "> Losses = ['0.0439', '0.0261', '0.0348', '0.0455']\n",
      "> Accuracies = ['98.81%', '99.38%', '99.07%', '98.33%']\n",
      "[LOG] Epoch = 0700\n",
      "> Losses = ['0.0423', '0.0247', '0.0333', '0.0434']\n",
      "> Accuracies = ['98.81%', '99.38%', '99.07%', '98.33%']\n",
      "[LOG] Epoch = 0800\n",
      "> Losses = ['0.0410', '0.0237', '0.0321', '0.0417']\n",
      "> Accuracies = ['98.81%', '99.38%', '99.07%', '98.33%']\n",
      "[LOG] Epoch = 0900\n",
      "> Losses = ['0.0400', '0.0228', '0.0311', '0.0403']\n",
      "> Accuracies = ['98.81%', '99.59%', '99.17%', '98.61%']\n",
      "[LOG] Epoch = 1000\n",
      "> Losses = ['0.0391', '0.0221', '0.0302', '0.0391']\n",
      "> Accuracies = ['98.81%', '99.59%', '99.17%', '98.61%']\n"
     ]
    }
   ],
   "source": [
    "losses, accuracies = federated_learning(clients)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3fae5f41",
   "metadata": {},
   "source": [
    "# Virtualize record of training process"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "a7530af2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "def plot_federated_graphs(diagnosis_title, losses, accuracies):\n",
    "    for i in range(n_hospitals):\n",
    "        plt.plot(losses[i], label=f'Hospital {i+1}')\n",
    "    legend = plt.legend(loc='upper right', shadow=True)\n",
    "    plt.title(f\"{diagnosis_title} - Training Loss\")\n",
    "    plt.xlabel(\"Iterations\")\n",
    "    plt.ylabel(\"Training Loss\")\n",
    "    plt.show()\n",
    "    for i in range(n_hospitals):\n",
    "        plt.plot(accuracies[i], label=f'Hospital {i+1}')\n",
    "    legend = plt.legend(loc='lower right', shadow=True)\n",
    "    plt.title(f\"{diagnosis_title} - Training Accuracy\")\n",
    "    plt.xlabel(\"Iterations\")\n",
    "    plt.ylabel(\"Training Accuracy (Percent %)\")\n",
    "    plt.show()\n",
    "\n",
    "plot_federated_graphs('Breast cancer', losses, accuracies)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "da465fc9",
   "metadata": {},
   "source": [
    "# Model parameters after training proccess"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "50bff47a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Model parameters:\n",
      "  | Weights: Parameter containing:\n",
      "tensor([[-0.7683,  0.7610,  0.0975,  0.7507,  0.5690,  0.1633,  0.0822,  0.5890,\n",
      "          0.7022, -0.0181, -0.6341,  0.6140, -0.3745,  0.4600,  0.6199, -0.0974,\n",
      "         -0.0828,  0.0506,  0.3189, -0.3061, -0.4902,  0.8851,  0.3885,  0.9437,\n",
      "          0.7826,  0.1689,  0.5525,  0.2612,  0.6942,  0.5343,  0.2268]],\n",
      "       requires_grad=True)\n",
      "  | Bias: Parameter containing:\n",
      "tensor(0.7911, requires_grad=True)\n"
     ]
    }
   ],
   "source": [
    "clients[0].decrypted_model_params()\n",
    "global_model = clients[0].local_model\n",
    "\n",
    "print('\\nModel parameters:')\n",
    "print('  | Weights: %s' % global_model.linear.weight) ### Virtualize record of training processlobal_model.linear.weight)\n",
    "print('  | Bias: %s' % global_model.linear.bias)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "db7d8e6f",
   "metadata": {},
   "source": [
    "#  Evaluating the Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "b8822525",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Testing Accuracy = 89.80%\n"
     ]
    }
   ],
   "source": [
    "# prepare data for testing model\n",
    "df_test = pd.read_csv('data/test.csv')\n",
    "df_test[\"diagnostic\"] = (df_test[\"diagnostic\"] == \"M\").astype(int)\n",
    "test , X_test , Y_test  = scale_dataset(df_test , False)\n",
    "\n",
    "test_acc = compute_federated_accuracy(global_model, X_test, Y_test)\n",
    "print('\\nTesting Accuracy = {}'.format(to_percent(test_acc)))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}