multilayer_perceptron_from_scratch.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Многослойный перцептрон с нуля"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Содержание\n",
    "\n",
    "1. Многослойный перцептрон с двумя параметрами на iris.csv  \n",
    "2. Многослойный перцептрон с четырьмя параметрами на iris.csv"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Многослойный перцептрон с двумя параметрами на iris.csv "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Предобработка данных"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt # для визуализации\n",
    "import os # для обработки пути к данным как на Unix-системах, так и на Windows"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(42) # фиксирование псевдорандомного генератора"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "DATA_DIR = \"data\" # путь к данным"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "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>sepal_length</th>\n",
       "      <th>sepal_width</th>\n",
       "      <th>petal_length</th>\n",
       "      <th>petal_width</th>\n",
       "      <th>species</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6.1</td>\n",
       "      <td>2.8</td>\n",
       "      <td>4.7</td>\n",
       "      <td>1.2</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5.7</td>\n",
       "      <td>3.8</td>\n",
       "      <td>1.7</td>\n",
       "      <td>0.3</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>7.7</td>\n",
       "      <td>2.6</td>\n",
       "      <td>6.9</td>\n",
       "      <td>2.3</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>6.0</td>\n",
       "      <td>2.9</td>\n",
       "      <td>4.5</td>\n",
       "      <td>1.5</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6.8</td>\n",
       "      <td>2.8</td>\n",
       "      <td>4.8</td>\n",
       "      <td>1.4</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   sepal_length  sepal_width  petal_length  petal_width  species\n",
       "0           6.1          2.8           4.7          1.2        1\n",
       "1           5.7          3.8           1.7          0.3        0\n",
       "2           7.7          2.6           6.9          2.3        2\n",
       "3           6.0          2.9           4.5          1.5        1\n",
       "4           6.8          2.8           4.8          1.4        1"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_iris = pd.read_csv(os.path.join(DATA_DIR, 'iris.csv')) # прочитать данные из CSV\n",
    "df_iris = df_iris.sample(frac=1, random_state=42) # перемешать данные\n",
    "df_iris = df_iris.reset_index(drop=True) # сбросить индексы для красоты\n",
    "\n",
    "# можно было записать одной строкой\n",
    "# df_iris = pd.read_csv(os.path.join(DATA_DIR, 'iris.csv')).sample(frac=1, random_state=42).reset_index(drop=True)\n",
    "\n",
    "df_iris['species'] = df_iris['species'].astype('category').cat.codes # заменить setosa и versicolor на 0 и 1\n",
    "df_iris.head() # вывести первые 5 рядов"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X_train.shape: (67, 2)\n",
      "y_train.shape: (67,)\n",
      "X_test.shape: (83, 2)\n",
      "y_test.shape: (83,)\n"
     ]
    }
   ],
   "source": [
    "# тренировочный датасет\n",
    "X_train = df_iris[['petal_length', 'petal_width']][:67].values # два параметра - petal_length и petal_width\n",
    "y_train = df_iris['species'][:67].values\n",
    "\n",
    "# тестовый датасет\n",
    "X_test = df_iris[['petal_length', 'petal_width']][67:].values # два параметра - petal_length и petal_width\n",
    "y_test = df_iris['species'][67:].values\n",
    "\n",
    "print(\"X_train.shape: {}\".format(X_train.shape))\n",
    "print(\"y_train.shape: {}\".format(y_train.shape))\n",
    "print(\"X_test.shape: {}\".format(X_test.shape))\n",
    "print(\"y_test.shape: {}\".format(y_test.shape))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Визуализация данных"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(X_train[y_train == 0][:, 0], X_train[y_train == 0][:, 1], 'rx', label='Train - Setosa')\n",
    "\n",
    "plt.plot(X_train[y_train == 1][:, 0], X_train[y_train == 1][:, 1], 'bx', label='Train - Versicolor')\n",
    "\n",
    "plt.plot(X_train[y_train == 2][:, 0], X_train[y_train == 2][:, 1], 'gx', label='Train - Virginica')\n",
    "\n",
    "plt.xlabel(\"Petal length\") # надпись по оси X\n",
    "plt.ylabel(\"Petal width\") # надпись по оси Y\n",
    "plt.legend() # отображение легенды\n",
    "plt.show() # отображение графика"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Создание и тренировка модели"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "########################################## TODO ########################################################"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Оценка модели"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "########################################## TODO ########################################################"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Визуализация работы модели"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "h = 0.01\n",
    "x_min, x_max = X_train[:, 0].min()-1, X_train[:, 0].max()+1\n",
    "y_min, y_max = X_train[:, 1].min()-1, X_train[:, 1].max()+1\n",
    "xx, yy = np.meshgrid(np.arange(x_min, x_max, h),\n",
    "                     np.arange(y_min, y_max, h))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "Z = clf.predict(np.c_[xx.ravel(), yy.ravel()])\n",
    "Z = Z.reshape(xx.shape)\n",
    "plt.plot(X_train[y_train == 0][:, 0], X_train[y_train == 0][:, 1], 'rx', label='Train')\n",
    "plt.plot(X_train[y_train == 1][:, 0], X_train[y_train == 1][:, 1], 'bx')\n",
    "plt.plot(X_train[y_train == 2][:, 0], X_train[y_train == 2][:, 1], 'bx')\n",
    "plt.plot(X_test[y_test == 0][:, 0], X_test[y_test == 0][:, 1], 'ro', label='Test')\n",
    "plt.plot(X_test[y_test == 1][:, 0], X_test[y_test == 1][:, 1], 'bo')\n",
    "plt.plot(X_test[y_test == 2][:, 0], X_test[y_test == 2][:, 1], 'go')\n",
    "plt.contourf(xx, yy, Z, cmap=plt.cm.RdBu, alpha=0.1) # цветовые области\n",
    "\n",
    "plt.xlabel(\"Petal length\") # надпись по оси X\n",
    "plt.ylabel(\"Petal width\") # надпись по оси Y\n",
    "plt.legend() # отображение легенды\n",
    "plt.show() # отображение графика"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Многослойный перцептрон с четырьмя параметрами на iris.csv "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Предобработка данных"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import os # для обработки пути к данным как на Unix-системах, так и на Windows"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(42) # фиксирование псевдорандомного генератора"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "DATA_DIR = \"data\" # путь к данным"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "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>sepal_length</th>\n",
       "      <th>sepal_width</th>\n",
       "      <th>petal_length</th>\n",
       "      <th>petal_width</th>\n",
       "      <th>species</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6.1</td>\n",
       "      <td>2.8</td>\n",
       "      <td>4.7</td>\n",
       "      <td>1.2</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5.7</td>\n",
       "      <td>3.8</td>\n",
       "      <td>1.7</td>\n",
       "      <td>0.3</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>7.7</td>\n",
       "      <td>2.6</td>\n",
       "      <td>6.9</td>\n",
       "      <td>2.3</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>6.0</td>\n",
       "      <td>2.9</td>\n",
       "      <td>4.5</td>\n",
       "      <td>1.5</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6.8</td>\n",
       "      <td>2.8</td>\n",
       "      <td>4.8</td>\n",
       "      <td>1.4</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   sepal_length  sepal_width  petal_length  petal_width  species\n",
       "0           6.1          2.8           4.7          1.2        1\n",
       "1           5.7          3.8           1.7          0.3        0\n",
       "2           7.7          2.6           6.9          2.3        2\n",
       "3           6.0          2.9           4.5          1.5        1\n",
       "4           6.8          2.8           4.8          1.4        1"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_iris = pd.read_csv(os.path.join(DATA_DIR, 'iris.csv')) # прочитать данные из CSV\n",
    "df_iris = df_iris.sample(frac=1, random_state=42) # перемешать данные\n",
    "df_iris = df_iris.reset_index(drop=True) # сбросить индексы для красоты\n",
    "\n",
    "# можно было записать одной строкой\n",
    "# df_iris = pd.read_csv(os.path.join(DATA_DIR, 'iris.csv')).sample(frac=1, random_state=42).reset_index(drop=True)\n",
    "\n",
    "df_iris['species'] = df_iris['species'].astype('category').cat.codes # заменить setosa и versicolor на 0 и 1\n",
    "df_iris.head() # вывести первые 5 рядов"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X_train.shape: (67, 4)\n",
      "y_train.shape: (67,)\n",
      "X_test.shape: (83, 4)\n",
      "y_test.shape: (83,)\n"
     ]
    }
   ],
   "source": [
    "# тренировочный датасет\n",
    "X_train = df_iris.drop(['species'], axis=1).iloc[:67].values # четыре параметра\n",
    "y_train = df_iris['species'][:67].values\n",
    "\n",
    "# тестовый датасет\n",
    "X_test = df_iris.drop(['species'], axis=1).iloc[67:].values # четыре параметра\n",
    "y_test = df_iris['species'][67:].values\n",
    "\n",
    "print(\"X_train.shape: {}\".format(X_train.shape))\n",
    "print(\"y_train.shape: {}\".format(y_train.shape))\n",
    "print(\"X_test.shape: {}\".format(X_test.shape))\n",
    "print(\"y_test.shape: {}\".format(y_test.shape))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Создание и тренировка модели"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "########################################## TODO ########################################################"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Оценка модели"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "########################################## TODO ########################################################"
   ]
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}