CustomNNClassifier.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Basic ML Solutions\n",
    "# Classification Algorithim using Custom NN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "import torch "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Serial No.</th>\n",
       "      <th>GRE Score</th>\n",
       "      <th>TOEFL Score</th>\n",
       "      <th>University Rating</th>\n",
       "      <th>SOP</th>\n",
       "      <th>LOR</th>\n",
       "      <th>CGPA</th>\n",
       "      <th>Research</th>\n",
       "      <th>Chance of Admit</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>262</th>\n",
       "      <td>263</td>\n",
       "      <td>308</td>\n",
       "      <td>103</td>\n",
       "      <td>2</td>\n",
       "      <td>2.5</td>\n",
       "      <td>4.0</td>\n",
       "      <td>8.36</td>\n",
       "      <td>1</td>\n",
       "      <td>0.70</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>240</th>\n",
       "      <td>241</td>\n",
       "      <td>296</td>\n",
       "      <td>101</td>\n",
       "      <td>1</td>\n",
       "      <td>2.5</td>\n",
       "      <td>3.0</td>\n",
       "      <td>7.68</td>\n",
       "      <td>0</td>\n",
       "      <td>0.60</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>383</th>\n",
       "      <td>384</td>\n",
       "      <td>300</td>\n",
       "      <td>100</td>\n",
       "      <td>3</td>\n",
       "      <td>3.0</td>\n",
       "      <td>3.5</td>\n",
       "      <td>8.26</td>\n",
       "      <td>0</td>\n",
       "      <td>0.62</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>385</th>\n",
       "      <td>386</td>\n",
       "      <td>335</td>\n",
       "      <td>117</td>\n",
       "      <td>5</td>\n",
       "      <td>5.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>9.82</td>\n",
       "      <td>1</td>\n",
       "      <td>0.96</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>77</th>\n",
       "      <td>78</td>\n",
       "      <td>301</td>\n",
       "      <td>99</td>\n",
       "      <td>2</td>\n",
       "      <td>3.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>8.22</td>\n",
       "      <td>0</td>\n",
       "      <td>0.64</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     Serial No.  GRE Score  TOEFL Score  University Rating  SOP  LOR   CGPA  \\\n",
       "262         263        308          103                  2  2.5   4.0  8.36   \n",
       "240         241        296          101                  1  2.5   3.0  7.68   \n",
       "383         384        300          100                  3  3.0   3.5  8.26   \n",
       "385         386        335          117                  5  5.0   5.0  9.82   \n",
       "77           78        301           99                  2  3.0   2.0  8.22   \n",
       "\n",
       "     Research  Chance of Admit   \n",
       "262         1              0.70  \n",
       "240         0              0.60  \n",
       "383         0              0.62  \n",
       "385         1              0.96  \n",
       "77          0              0.64  "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "admit_data = pd.read_csv('admission_predict.csv')\n",
    "admit_data.sample(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>GRE Score</th>\n",
       "      <th>TOEFL Score</th>\n",
       "      <th>University Rating</th>\n",
       "      <th>SOP</th>\n",
       "      <th>LOR</th>\n",
       "      <th>CGPA</th>\n",
       "      <th>Research</th>\n",
       "      <th>Chance of Admit</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>400.000000</td>\n",
       "      <td>400.000000</td>\n",
       "      <td>400.000000</td>\n",
       "      <td>400.000000</td>\n",
       "      <td>400.000000</td>\n",
       "      <td>400.000000</td>\n",
       "      <td>400.000000</td>\n",
       "      <td>400.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>316.807500</td>\n",
       "      <td>107.410000</td>\n",
       "      <td>3.087500</td>\n",
       "      <td>3.400000</td>\n",
       "      <td>3.452500</td>\n",
       "      <td>8.598925</td>\n",
       "      <td>0.547500</td>\n",
       "      <td>0.724350</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>11.473646</td>\n",
       "      <td>6.069514</td>\n",
       "      <td>1.143728</td>\n",
       "      <td>1.006869</td>\n",
       "      <td>0.898478</td>\n",
       "      <td>0.596317</td>\n",
       "      <td>0.498362</td>\n",
       "      <td>0.142609</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>290.000000</td>\n",
       "      <td>92.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>6.800000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.340000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>308.000000</td>\n",
       "      <td>103.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>2.500000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>8.170000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.640000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>317.000000</td>\n",
       "      <td>107.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>3.500000</td>\n",
       "      <td>3.500000</td>\n",
       "      <td>8.610000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.730000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>325.000000</td>\n",
       "      <td>112.000000</td>\n",
       "      <td>4.000000</td>\n",
       "      <td>4.000000</td>\n",
       "      <td>4.000000</td>\n",
       "      <td>9.062500</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.830000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>340.000000</td>\n",
       "      <td>120.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>9.920000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.970000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        GRE Score  TOEFL Score  University Rating         SOP        LOR   \\\n",
       "count  400.000000   400.000000         400.000000  400.000000  400.000000   \n",
       "mean   316.807500   107.410000           3.087500    3.400000    3.452500   \n",
       "std     11.473646     6.069514           1.143728    1.006869    0.898478   \n",
       "min    290.000000    92.000000           1.000000    1.000000    1.000000   \n",
       "25%    308.000000   103.000000           2.000000    2.500000    3.000000   \n",
       "50%    317.000000   107.000000           3.000000    3.500000    3.500000   \n",
       "75%    325.000000   112.000000           4.000000    4.000000    4.000000   \n",
       "max    340.000000   120.000000           5.000000    5.000000    5.000000   \n",
       "\n",
       "             CGPA    Research  Chance of Admit   \n",
       "count  400.000000  400.000000        400.000000  \n",
       "mean     8.598925    0.547500          0.724350  \n",
       "std      0.596317    0.498362          0.142609  \n",
       "min      6.800000    0.000000          0.340000  \n",
       "25%      8.170000    0.000000          0.640000  \n",
       "50%      8.610000    1.000000          0.730000  \n",
       "75%      9.062500    1.000000          0.830000  \n",
       "max      9.920000    1.000000          0.970000  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# drop Serial #\n",
    "admit_data = admit_data.drop(columns=['Serial No.'], axis=1)\n",
    "admit_data.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# need to standardize the data, as mean and std are varying a lot\n",
    "# std should be close to 1\n",
    "from sklearn.preprocessing import scale\n",
    "admit_data[['GRE Score', 'TOEFL Score', 'SOP', 'LOR ', 'CGPA']] = scale(\n",
    "    admit_data[['GRE Score', 'TOEFL Score', 'SOP', 'LOR ', 'CGPA']])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>GRE Score</th>\n",
       "      <th>TOEFL Score</th>\n",
       "      <th>University Rating</th>\n",
       "      <th>SOP</th>\n",
       "      <th>LOR</th>\n",
       "      <th>CGPA</th>\n",
       "      <th>Research</th>\n",
       "      <th>Chance of Admit</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>4.000000e+02</td>\n",
       "      <td>4.000000e+02</td>\n",
       "      <td>400.000000</td>\n",
       "      <td>4.000000e+02</td>\n",
       "      <td>4.000000e+02</td>\n",
       "      <td>4.000000e+02</td>\n",
       "      <td>400.000000</td>\n",
       "      <td>400.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>-3.785861e-16</td>\n",
       "      <td>5.412337e-16</td>\n",
       "      <td>3.087500</td>\n",
       "      <td>1.859624e-16</td>\n",
       "      <td>-3.019807e-16</td>\n",
       "      <td>8.076873e-16</td>\n",
       "      <td>0.547500</td>\n",
       "      <td>0.724350</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>1.001252e+00</td>\n",
       "      <td>1.001252e+00</td>\n",
       "      <td>1.143728</td>\n",
       "      <td>1.001252e+00</td>\n",
       "      <td>1.001252e+00</td>\n",
       "      <td>1.001252e+00</td>\n",
       "      <td>0.498362</td>\n",
       "      <td>0.142609</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>-2.339367e+00</td>\n",
       "      <td>-2.542098e+00</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-2.386613e+00</td>\n",
       "      <td>-2.733036e+00</td>\n",
       "      <td>-3.020504e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.340000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>-7.685900e-01</td>\n",
       "      <td>-7.274920e-01</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>-8.949798e-01</td>\n",
       "      <td>-5.042604e-01</td>\n",
       "      <td>-7.201909e-01</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.640000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>1.679859e-02</td>\n",
       "      <td>-6.763531e-02</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>9.944220e-02</td>\n",
       "      <td>5.293342e-02</td>\n",
       "      <td>1.859559e-02</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.730000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>7.149218e-01</td>\n",
       "      <td>7.571856e-01</td>\n",
       "      <td>4.000000</td>\n",
       "      <td>5.966532e-01</td>\n",
       "      <td>6.101273e-01</td>\n",
       "      <td>7.783704e-01</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.830000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>2.023903e+00</td>\n",
       "      <td>2.076899e+00</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>1.591075e+00</td>\n",
       "      <td>1.724515e+00</td>\n",
       "      <td>2.218165e+00</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.970000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          GRE Score   TOEFL Score  University Rating           SOP  \\\n",
       "count  4.000000e+02  4.000000e+02         400.000000  4.000000e+02   \n",
       "mean  -3.785861e-16  5.412337e-16           3.087500  1.859624e-16   \n",
       "std    1.001252e+00  1.001252e+00           1.143728  1.001252e+00   \n",
       "min   -2.339367e+00 -2.542098e+00           1.000000 -2.386613e+00   \n",
       "25%   -7.685900e-01 -7.274920e-01           2.000000 -8.949798e-01   \n",
       "50%    1.679859e-02 -6.763531e-02           3.000000  9.944220e-02   \n",
       "75%    7.149218e-01  7.571856e-01           4.000000  5.966532e-01   \n",
       "max    2.023903e+00  2.076899e+00           5.000000  1.591075e+00   \n",
       "\n",
       "               LOR           CGPA    Research  Chance of Admit   \n",
       "count  4.000000e+02  4.000000e+02  400.000000        400.000000  \n",
       "mean  -3.019807e-16  8.076873e-16    0.547500          0.724350  \n",
       "std    1.001252e+00  1.001252e+00    0.498362          0.142609  \n",
       "min   -2.733036e+00 -3.020504e+00    0.000000          0.340000  \n",
       "25%   -5.042604e-01 -7.201909e-01    0.000000          0.640000  \n",
       "50%    5.293342e-02  1.859559e-02    1.000000          0.730000  \n",
       "75%    6.101273e-01  7.783704e-01    1.000000          0.830000  \n",
       "max    1.724515e+00  2.218165e+00    1.000000          0.970000  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "admit_data.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "x = admit_data.drop('Chance of Admit ', axis=1)\n",
    "y = admit_data[['Chance of Admit ']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Chance of Admit</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.92</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.76</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.72</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.80</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.65</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Chance of Admit \n",
       "0              0.92\n",
       "1              0.76\n",
       "2              0.72\n",
       "3              0.80\n",
       "4              0.65"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "target = y.copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "target.replace(to_replace = y[y >= 0.80], value=int(2), inplace=True)\n",
    "target.replace(to_replace = y[y >= 0.60], value=int(1), inplace=True)\n",
    "target.replace(to_replace = y[y < 0.60], value=int(0), inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "y = target"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([2., 1., 0.])"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y['Chance of Admit '].unique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.2)\n",
    "\n",
    "x_train = torch.tensor(x_train.values, dtype=torch.float)\n",
    "x_test = torch.tensor(x_test.values, dtype=torch.float)\n",
    "y_train = torch.from_numpy(y_train.values).view(-1).long()\n",
    "y_test = torch.from_numpy(y_test.values).view(-1).long()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# instead of using Sequential Models, build Custom Model\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "inp = x_train.shape[1]\n",
    "out = 3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create a class called Net, and this Net class inherits from the nn.Module class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Net(nn.Module):\n",
    "    \n",
    "    def __init__(self, hidden_size, activation_fn='relu', apply_dropout=False):\n",
    "        super(Net, self).__init__()\n",
    "        \n",
    "        self.fc1 = nn.Linear(inp, hidden_size)\n",
    "        self.fc2 = nn.Linear(hidden_size, hidden_size)\n",
    "        self.fc3 = nn.Linear(hidden_size, out)\n",
    "        \n",
    "        self.hidden_size = hidden_size\n",
    "        self.activation_fn = activation_fn\n",
    "        \n",
    "        self.dropout = None\n",
    "        if apply_dropout:\n",
    "            self.dropout = nn.Dropout(p=0.2)\n",
    "            \n",
    "    def forward(self, x):\n",
    "        \n",
    "        activation_fn = None\n",
    "        if self.activation_fn == 'sigmoid':\n",
    "            activation_fn = F.torch.sigmoid\n",
    "        if self.activation_fn == 'tanh':\n",
    "            activation_fn = F.torch.tanh\n",
    "        if self.activation_fn == 'relu':\n",
    "            activation_fn = F.torch.relu\n",
    "            \n",
    "        x = activation_fn(self.fc1(x))\n",
    "        x = activation_fn(self.fc2(x))\n",
    "        \n",
    "        if self.dropout != None:\n",
    "            x = self.dropout(x)\n",
    "        \n",
    "        x = self.fc3(x)\n",
    "        return F.log_softmax(x, dim=-1)       "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_evaluate_model(model, learning_rate=0.001):\n",
    "    epoch_data = []\n",
    "    epochs = 1001\n",
    "    \n",
    "    optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)\n",
    "    \n",
    "    loss_fn = nn.NLLLoss()\n",
    "    \n",
    "    test_accuracy=0.0\n",
    "    for epoch in range(1, epochs):\n",
    "        \n",
    "        optimizer.zero_grad()\n",
    "        \n",
    "        # setting the model to training mode\n",
    "        model.train()\n",
    "        \n",
    "        y_pred = model(x_train)\n",
    "        \n",
    "        loss = loss_fn(y_pred, y_train)\n",
    "        loss.backward()\n",
    "        \n",
    "        optimizer.step()\n",
    "        \n",
    "        # setting the model to evaluation mode\n",
    "        model.eval()\n",
    "        \n",
    "        y_pred_tst = model(x_test)\n",
    "        loss_tst = loss_fn(y_pred_tst, y_test)\n",
    "        \n",
    "        _, pred = y_pred_tst.data.max(1)\n",
    "        \n",
    "        test_accuracy = pred.eq(y_test.data).sum().item() / 80 #y_test.values.size\n",
    "        \n",
    "        epoch_data.append([epoch, loss.data.item(), loss_tst.data.item(), test_accuracy])\n",
    "        \n",
    "        if epoch % 200 == 0:\n",
    "            print('epoch - %d (%d%%) train loss - %.2f test loss - %.2f Test Accuracy - %.4f'\n",
    "                 % (epoch, epoch/15, loss.data.item(), loss_tst.data.item(), test_accuracy))\n",
    "        \n",
    "        \n",
    "    return {\n",
    "        'model': model,\n",
    "        'epoch_data': epoch_data,\n",
    "        '# epochs': epochs,\n",
    "        'optimizer': optimizer,\n",
    "        'loss_fn': loss_fn, \n",
    "        'test accuracy': test_accuracy,\n",
    "        '_, pred': y_pred_tst.data.max(1),\n",
    "        'actual ypred label': y_test\n",
    "    }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Net(\n",
       "  (fc1): Linear(in_features=7, out_features=3, bias=True)\n",
       "  (fc2): Linear(in_features=3, out_features=3, bias=True)\n",
       "  (fc3): Linear(in_features=3, out_features=3, bias=True)\n",
       ")"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "net = Net(hidden_size=3, activation_fn='sigmoid', apply_dropout=False)\n",
    "net"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch - 200 (13%) train loss - 1.03 test loss - 1.02 Test Accuracy - 0.5000\n",
      "epoch - 400 (26%) train loss - 0.98 test loss - 0.98 Test Accuracy - 0.5000\n",
      "epoch - 600 (40%) train loss - 0.86 test loss - 0.87 Test Accuracy - 0.7000\n",
      "epoch - 800 (53%) train loss - 0.73 test loss - 0.77 Test Accuracy - 0.7375\n",
      "epoch - 1000 (66%) train loss - 0.66 test loss - 0.72 Test Accuracy - 0.7125\n"
     ]
    }
   ],
   "source": [
    "res_3_sigmoid = train_evaluate_model(net)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Net(\n",
       "  (fc1): Linear(in_features=7, out_features=3, bias=True)\n",
       "  (fc2): Linear(in_features=3, out_features=3, bias=True)\n",
       "  (fc3): Linear(in_features=3, out_features=3, bias=True)\n",
       "  (dropout): Dropout(p=0.2, inplace=False)\n",
       ")"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "net = Net(hidden_size=3, activation_fn='sigmoid', apply_dropout=True)\n",
    "net"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch - 200 (13%) train loss - 1.03 test loss - 1.01 Test Accuracy - 0.5000\n",
      "epoch - 400 (26%) train loss - 0.99 test loss - 0.99 Test Accuracy - 0.5000\n",
      "epoch - 600 (40%) train loss - 0.92 test loss - 0.92 Test Accuracy - 0.6250\n",
      "epoch - 800 (53%) train loss - 0.82 test loss - 0.82 Test Accuracy - 0.7250\n",
      "epoch - 1000 (66%) train loss - 0.73 test loss - 0.74 Test Accuracy - 0.7250\n"
     ]
    }
   ],
   "source": [
    "res_3_sigmoid_drop = train_evaluate_model(net)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Net(\n",
       "  (fc1): Linear(in_features=7, out_features=10, bias=True)\n",
       "  (fc2): Linear(in_features=10, out_features=10, bias=True)\n",
       "  (fc3): Linear(in_features=10, out_features=3, bias=True)\n",
       ")"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "net = Net(hidden_size=10, activation_fn='tanh', apply_dropout=False)\n",
    "net"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch - 200 (13%) train loss - 0.49 test loss - 0.56 Test Accuracy - 0.7875\n",
      "epoch - 400 (26%) train loss - 0.42 test loss - 0.55 Test Accuracy - 0.7750\n",
      "epoch - 600 (40%) train loss - 0.38 test loss - 0.60 Test Accuracy - 0.7375\n",
      "epoch - 800 (53%) train loss - 0.36 test loss - 0.64 Test Accuracy - 0.7125\n",
      "epoch - 1000 (66%) train loss - 0.35 test loss - 0.66 Test Accuracy - 0.7250\n"
     ]
    }
   ],
   "source": [
    "res_10_tanh = train_evaluate_model(net)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Net(\n",
       "  (fc1): Linear(in_features=7, out_features=10, bias=True)\n",
       "  (fc2): Linear(in_features=10, out_features=10, bias=True)\n",
       "  (fc3): Linear(in_features=10, out_features=3, bias=True)\n",
       ")"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "net = Net(hidden_size=10, activation_fn='sigmoid', apply_dropout=False)\n",
    "net"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch - 200 (13%) train loss - 0.91 test loss - 0.91 Test Accuracy - 0.5000\n",
      "epoch - 400 (26%) train loss - 0.61 test loss - 0.67 Test Accuracy - 0.7000\n",
      "epoch - 600 (40%) train loss - 0.54 test loss - 0.63 Test Accuracy - 0.6875\n",
      "epoch - 800 (53%) train loss - 0.49 test loss - 0.61 Test Accuracy - 0.7375\n",
      "epoch - 1000 (66%) train loss - 0.45 test loss - 0.58 Test Accuracy - 0.7750\n"
     ]
    }
   ],
   "source": [
    "res_10_sigmoid = train_evaluate_model(net)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Net(\n",
       "  (fc1): Linear(in_features=7, out_features=25, bias=True)\n",
       "  (fc2): Linear(in_features=25, out_features=25, bias=True)\n",
       "  (fc3): Linear(in_features=25, out_features=3, bias=True)\n",
       ")"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "net = Net(hidden_size=25, activation_fn='sigmoid', apply_dropout=False)\n",
    "net"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch - 200 (13%) train loss - 0.55 test loss - 0.61 Test Accuracy - 0.7125\n",
      "epoch - 400 (26%) train loss - 0.45 test loss - 0.56 Test Accuracy - 0.7875\n",
      "epoch - 600 (40%) train loss - 0.41 test loss - 0.56 Test Accuracy - 0.7500\n",
      "epoch - 800 (53%) train loss - 0.38 test loss - 0.58 Test Accuracy - 0.7375\n",
      "epoch - 1000 (66%) train loss - 0.37 test loss - 0.60 Test Accuracy - 0.7250\n"
     ]
    }
   ],
   "source": [
    "res_25_sigmoid_drop = train_evaluate_model(net) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "result_model = res_10_sigmoid"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_epochs_data = pd.DataFrame(result_model['epoch_data'], columns=['epoch', 'train_loss', 'test_loss', 'accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>test_loss</th>\n",
       "      <th>accuracy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>1.116634</td>\n",
       "      <td>1.115349</td>\n",
       "      <td>0.175</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>1.113602</td>\n",
       "      <td>1.112247</td>\n",
       "      <td>0.175</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>1.110625</td>\n",
       "      <td>1.109199</td>\n",
       "      <td>0.275</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>1.107702</td>\n",
       "      <td>1.106209</td>\n",
       "      <td>0.400</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>1.104835</td>\n",
       "      <td>1.103274</td>\n",
       "      <td>0.475</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   epoch  train_loss  test_loss  accuracy\n",
       "0      1    1.116634   1.115349     0.175\n",
       "1      2    1.113602   1.112247     0.175\n",
       "2      3    1.110625   1.109199     0.275\n",
       "3      4    1.107702   1.106209     0.400\n",
       "4      5    1.104835   1.103274     0.475"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_epochs_data.head(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x720 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "f, (ax1, ax2) = plt.subplots(1,2, figsize=(15,10))\n",
    "df_epochs_data[['train_loss', 'test_loss']].plot(ax=ax1)\n",
    "df_epochs_data[['accuracy']].plot(ax=ax2)\n",
    "plt.ylim(bottom=0.5)\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}