hw8_part1_implement_sgd.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<center>\n",
    "<img src=\"../../img/ods_stickers.jpg\">\n",
    "## Открытый курс по машинному обучению\n",
    "</center>\n",
    "<center>Автор материала: программист-исследователь Mail.ru Group, старший преподаватель <br>Факультета Компьютерных Наук ВШЭ Юрий Кашницкий"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# <center> Домашнее задание № 8. Часть 1\n",
    "## <center> Реализация алгоритмов онлайн-обучения"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "В первой части задания вам предлагается реализовать два алгоритма – регрессор и классификатор, обучаемые стохастическим градиентным спуском (Stochastic Gradient Descent, SGD). [Веб-форма](https://goo.gl/forms/z8zENbMiaEAeB7nG3) для ответов."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## План 1 части домашнего задания\n",
    "    1.1. Линейная регрессия и SGD\n",
    "    1.2. Логистическая регрессия и SGD\n",
    "    1.3. Логистическая регрессия и SGDClassifier в задаче классификации отзывов к фильмам"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "В [статье](https://habrahabr.ru/company/ods/blog/326418/) было описано, как таким образом обучать регрессор, т.е. минимизировать квадратичную функцию потерь. Реализуем этот алгоритм."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Вначале проверим версии библиотек с помощью расширения [watermark](https://github.com/rasbt/watermark). Заодно и Vowpal Wabbit, который понадобится во 2 части задания – версия должна быть последняя на момент написания, 8.3.2."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#pip install watermark\n",
    "%load_ext watermark"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPython 3.5.2\n",
      "IPython 5.1.0\n",
      "\n",
      "numpy 1.12.1\n",
      "scipy 0.19.0\n",
      "pandas 0.19.2\n",
      "matplotlib 2.0.0\n",
      "sklearn 0.18.1\n",
      "\n",
      "compiler   : GCC 5.4.0 20160609\n",
      "system     : Linux\n",
      "release    : 4.4.0-75-generic\n",
      "machine    : x86_64\n",
      "processor  : x86_64\n",
      "CPU cores  : 4\n",
      "interpreter: 64bit\n"
     ]
    }
   ],
   "source": [
    "%watermark -v -m -p numpy,scipy,pandas,matplotlib,sklearn\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "8.3.2\r\n"
     ]
    }
   ],
   "source": [
    "!vw --version"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##  1.1. Линейная регрессия и SGD"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "from tqdm import tqdm\n",
    "from sklearn.base import BaseEstimator\n",
    "from sklearn.metrics import mean_squared_error, log_loss, roc_auc_score\n",
    "from sklearn.model_selection import train_test_split\n",
    "%matplotlib inline\n",
    "from matplotlib import pyplot as plt\n",
    "import seaborn as sns\n",
    "from sklearn.preprocessing import StandardScaler"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Реализуйте класс `OnlineRegressor`. Спецификация:\n",
    "- класс наследуется от `sklearn.base.BaseEstimator`\n",
    "- конструктор принимает параметры `eta` – шаг градиентного спуска (по умолчанию $10^{-3}$) и `n_iter` – число проходов по выборке (по умолчанию 10)\n",
    "- также в конструкторе должны создаваться списки `mse_` и `weights_` для отслеживания значений среднеквадратичной ошибки и вектора весов по итерациям градиентного спуска\n",
    "- Класс имеет методы `fit` и `predict`\n",
    "- Метод `fit` принимает матрицу `X` и вектор `y` (объекты `numpy.array`), добавляет к матрице `X` слева столбец из единиц, инициализирует вектор весов `w` **нулями** и в цикле с числом итераций `n_iter` обновляет веса (см. [статью](https://habrahabr.ru/company/ods/blog/326418/)), а также записывает получившиеся на данной итерации значения среднеквадратичной ошибки (именно MSE, SE слишком большими будут) и вектор весов `w` в предназначенные для этого списки. \n",
    "- В конце метод `fit` создает переменную `w_`, в которой хранится тот вектор весов, при котором ошибка минимальна\n",
    "- Метод `fit` должен возвращать текущий экземпляр класса `OnlineRegressor`, т.е. `self`\n",
    "- Метод `predict` принимает матрицу `X`, добавляет к ней слева столбец из единиц и возвращает вектор прогнозов модели, используя созданный методом `fit` вектор весов `w_`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "class OnlineRegressor(BaseEstimator):\n",
    "    \n",
    "    def __init__(self, eta=0.001, n_iter=10):\n",
    "        self.eta = eta\n",
    "        self.n_iter = n_iter\n",
    "        self.mse_ = []\n",
    "        self.weights = []\n",
    "        \n",
    "    def fit(self, X_train, y_train):\n",
    "        a = np.ones((X_train.shape[0],1))\n",
    "        X_train = np.concatenate((a, X_train), axis=1)\n",
    "        w = np.zeros(X_train.shape[1])\n",
    "        for t in range(self.n_iter):\n",
    "            for i in range(X_train.shape[0]):\n",
    "                w += self.eta * (y_train[i] - sum(w * X_train[i])) * X_train[i]\n",
    "                if (t>0) | (i>0):\n",
    "                    self.weights = np.append(self.weights, np.array(w))\n",
    "                else:\n",
    "                    self.weights = [np.array(w)]\n",
    "                self.mse_.append(mean_squared_error(y_train, np.cumsum((w*X_train), axis=1)[:,X_train.shape[1]-1]))\n",
    "        low_mse_index = self.mse_.index(min(self.mse_))\n",
    "        self.w_ = [self.weights[low_mse_index*X_train.shape[1]+i] for i in range(X_train.shape[1])]\n",
    "        return self\n",
    "        \n",
    "    def predict(self, X_test):\n",
    "        a = np.ones((X_test.shape[0],1))\n",
    "        X_test = np.concatenate((a, X_test), axis=1)\n",
    "        return np.cumsum((self.w_*X_test), axis=1)[:,X_test.shape[1]-1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.5, 0.5, 0.48524999999999996, 0.47574500999999997, 0.4759292437025, 0.47634168788447173, 0.46276049486049886, 0.45389493926649471, 0.45422317731968942, 0.45496528017060195, 0.44244516517226146, 0.43416588821203783, 0.43460417676848351, 0.4356054731668339, 0.42404953114392718, 0.41630828652285623, 0.41682808313428121, 0.41802877958007434, 0.40734977369717629, 0.40010273291798337]\n",
      "[ 0.          0.          0.          0.          0.          0.          0.01\n",
      "  0.          0.01        0.0199      0.          0.01        0.019701\n",
      " -0.000199    0.01        0.01940598 -0.00049402  0.00970498  0.02911487\n",
      " -0.00049402  0.01941387  0.03882372 -0.00049402  0.01941387  0.03844042\n",
      " -0.00087732  0.01941387  0.03787065 -0.00144709  0.0188441   0.04730351\n",
      " -0.00144709  0.02827695  0.05683047 -0.00144709  0.02827695  0.05627664\n",
      " -0.00200092  0.02827695  0.05545111 -0.00282645  0.02745143  0.06462209\n",
      " -0.00282645  0.0366224   0.07397587 -0.00282645  0.0366224   0.07326437\n",
      " -0.00353794  0.0366224   0.07220088 -0.00460143  0.03555891  0.08112329\n",
      " -0.00460143  0.04448131  0.09031205 -0.00460143  0.04448131]\n",
      "19\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([ 0.08110919,  0.11255271])"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAecAAAFKCAYAAAAnj5dkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f302d844048>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "test = OnlineRegressor(0.01, 5)\n",
    "test.weights\n",
    "test.fit(X_train, y_train)\n",
    "test.weights, test.w_\n",
    "plt.plot(test.mse_)\n",
    "test.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 1.  0.]\n",
      " [ 1.  1.]\n",
      " [ 0.  1.]\n",
      " [ 0.  0.]]\n",
      "[ 0.  0.]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([ 1.,  0.])"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train = np.array([[1., 0.], [1., 1.], [0., 1.], [0., 0.]])\n",
    "X_test = np.array([[2, 0], [0., 0.5]])\n",
    "y_train = np.array([0, 0, 1, 1])\n",
    "print(X_train)\n",
    "w = np.zeros(X_train.shape[1])\n",
    "print(w)\n",
    "mse_ = np.array([])\n",
    "mse_\n",
    "a = np.ones((X_train.shape[0],1))\n",
    "#X_train = np.concatenate((a, X_train), 1)\n",
    "w[0] += 1\n",
    "w"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Проверим работу алгоритма на данных по росту и весу. Будем прогнозировать рост (в дюймах) по весу (в фунтах)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "data_demo = pd.read_csv('weights_heights.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAe0AAAFYCAYAAAB+s6Q9AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nhBBCosT/Bz4TH2ONoCX/AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f302d844668>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(data_demo['Weight'], data_demo['Height']);\n",
    "plt.xlabel('Вес (фунты)')\n",
    "plt.ylabel('Рост (дюймы)');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X, y = data_demo['Weight'].values, data_demo['Height'].values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Выделим 70% под обучение, 30% – под проверку и масштабируем выборку."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X_train, X_valid, y_train, y_valid = train_test_split(X, y,\n",
    "                                                     test_size=0.3,\n",
    "                                                     random_state=17)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "scaler = StandardScaler()\n",
    "X_train_scaled = scaler.fit_transform(X_train.reshape([X_train.shape[0], 1]))\n",
    "X_valid_scaled = scaler.transform(X_valid.reshape([X_valid.shape[0], 1]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Обучите созданный вами `OnlineRegressor` на выборке `(X_train_scaled, y_train)`. Параметры оставьте по умолчанию."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "OnlineRegressor(eta=0.001, n_iter=10)"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf = OnlineRegressor()\n",
    "clf.fit(X_train_scaled, y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Изобразите на графике процесс обучения – как среднеквадратичная ошибка зависит от номера итерации стохастического градиентного спуска."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7f302b256320>]"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeoAAAFKCAYAAADScRzUAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ntYiIiIUpqEVERCxMQS0iImJh/x8eW6xjT/5BAAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f302b712630>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(clf.mse_[:2000])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Выведите наименьшее значение среднеквадратичной ошибки и лучший вектор весов модели."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.71513524066\n",
      "12626\n",
      "67.9898496959 0.944476053405\n"
     ]
    }
   ],
   "source": [
    "print(min(clf.mse_))\n",
    "print(clf.mse_.index(min(clf.mse_)))\n",
    "print(clf.weights[12626*2], clf.weights[12626*2+1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Постройте график того, как менялись значения весов модели ($w_0$ и $w_1$) по мере обучения."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7f302a9e5710>]"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAd8AAAFKCAYAAABcq1WoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nONn/BzCnXii1Oqv7AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f302b2a8400>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "w0 = [clf.weights[i] for i in range(len(clf.weights)) if i % 2 == 0]\n",
    "w1 = [clf.weights[i] for i in range(len(clf.weights)) if i % 2 == 1]\n",
    "plt.plot(w0[:2000])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7f302a955cc0>]"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAecAAAFKCAYAAAAnj5dkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nGc5CCCGEYmQ4CyGEEIqR4SyEEEIo5he2bqrkkeOkHAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f302b264198>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(w1[:2000])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Сделайте прогноз для отложенной выборки `(X_valid_scaled, y_valid)` и посмотрите на MSE."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2.6708681207033784"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred = clf.predict(X_valid_scaled)\n",
    "mean_squared_error(y_valid, pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Теперь следайте то же самое, но с `LinearRegression` из `sklearn.linear_model`. Посчитайте MSE для отложенной выборки."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LinearRegression(copy_X=True, fit_intercept=True, n_jobs=1, normalize=False)"
      ]
     },
     "execution_count": 137,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.linear_model import LinearRegression\n",
    "lr = LinearRegression()\n",
    "lr.fit(X_train_scaled, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2.6708307676676339"
      ]
     },
     "execution_count": 138,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred = lr.predict(X_valid_scaled)\n",
    "mean_squared_error(y_valid, pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "<font color='red'>Вопрос 1.</font> В каком знаке после разделителя отличаются MSE линейной регрессии и `OnlineRegressor` для отложенной выборки?\n",
    " - 2\n",
    " - 3\n",
    " - 4\n",
    " - 5"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1.2. Логистическая регрессия и SGD\n",
    "Теперь давайте разберемся, как при таком же стохастическом подходе обучать логистическую регрессию."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Задача классификации, $X$ – обучающая выборка размеров $\\ell \\times (d+1)$ (первый столбец – вектор из единиц), $y$ – вектор ответов, $y_i \\in \\{-1, 1\\}$.\n",
    "В [4 статье](https://habrahabr.ru/company/ods/blog/323890/) серии мы подробно разбирали, как логистическая регрессия с $L_2$-регуляризацией сводится к задаче минимизации:\n",
    "$$ C\\sum_{i=1}^\\ell \\log{(1 + e^{-y_iw^Tx_i})} + \\frac{1}{2}\\sum_{j=1}^d w_j^2 \\rightarrow min_w$$"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font color='red'>Вопрос 2.</font> По какой формуле будут пересчитываться веса логистической регрессии при обучении стохастическим градиентным спуском?\n",
    " - $w_j^{(t+1)} = w_j^{(t)} + \\eta (Cy_i x_{ij} \\sigma(y_iw^Tx_i) +  \\delta_{j\\neq0} w_j)$\n",
    " - $w_j^{(t+1)} = w_j^{(t)} - \\eta (Cy_i x_{ij} \\sigma(-y_iw^Tx_i) +  \\delta_{j\\neq0}w_j)$\n",
    " - $w_j^{(t+1)} = w_j^{(t)} - \\eta (Cy_i x_{ij} \\sigma(y_iw^Tx_i) -  \\delta_{j\\neq0}w_j )$\n",
    " - $w_j^{(t+1)} = w_j^{(t)} + \\eta (Cy_i x_{ij} \\sigma(-y_iw^Tx_i) -  \\delta_{j\\neq0}w_j)$\n",
    " \n",
    "Здесь \n",
    "- $i \\in {0,\\ldots, \\ell-1}, j \\in {0,\\ldots, d}$\n",
    "- C – коэффициент регуляризации\n",
    "- $x_{ij} $ – элемент матрицы X в строке $i$ и столбце $j$ (нумерация с 0), \n",
    "- $x_i$ – $i$-ая строка матрицы $X$ (нумерация с 0), \n",
    "- $w_j^{(t)}$ – значение $j$-ого элемента вектора весов $w$ на шаге $t$ стохастического градиентного спуска\n",
    "- $\\eta$ – небольшая константа, шаг градиентного спуска\n",
    "- $\\delta_{j\\neq0}$ – символ Кронекера, то есть 1, когда $j\\neq0$ и $0$ – в противном случае"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "возьмём производную и выкинем сумму. Получится --+ Но нам нужен шаг в сторону антиградиента +-- =>4"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Реализуйте класс `OnlineClassifier`. Спецификация:\n",
    "- класс наследуется от `sklearn.base.BaseEstimator`\n",
    "- конструктор принимает параметры `eta` – шаг градиентного спуска (по умолчанию $10^{-3}$), `n_iter` – число проходов по выборке (по умолчанию 10) и C – коэффициент регуляризации\n",
    "- также в конструкторе должны создаваться списки `loss_` и `weights_` для отслеживания значений логистических потерь и вектора весов по итерациям градиентного спуска\n",
    "- Класс имеет методы `fit`, `predict` и `predict_proba`\n",
    "- Метод `fit` принимает матрицу `X` и вектор `y` (объекты `numpy.array`, рассматриваем только случай бинарной классификации, и значения в векторе `y` могут быть -1 и 1), добавляет к матрице `X` слева столбец из единиц, инициализирует вектор весов `w` **нулями** и в цикле с числом итераций `n_iter` обновляет веса по выведенной вами формуле, а также записывает получившиеся на данной итерации значения log_loss и вектор весов `w` в предназначенные для этого списки. \n",
    "- В конце метод `fit` создает переменную `w_`, в которой хранится тот вектор весов, при котором ошибка минимальна\n",
    "- Метод `fit` должен возвращать текущий экземпляр класса `OnlineClassifier`, т.е. `self`\n",
    "- Метод `predict_proba` принимает матрицу `X`, добавляет к ней слева столбец из единиц и возвращает матрицу прогнозов модели (такую же, какую возвращают методы `predict_proba` моделей `sklearn`), используя созданный методом `fit` вектор весов `w_`\n",
    "- Метод `predict` вызывает метод  `predict_proba` и возвращает вектор ответов: -1, если предсказанная вероятность класса 1 меньше 0.5 и 1 – в противном случае\n",
    "- И еще **важный момент**: во избежание вычислительных проблем из-за слишком больших или малых значений под экспонентной (overflow & underflow) используйте написанную функцию `sigma`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def sigma(z):\n",
    "    z = z.flatten()\n",
    "    z[z > 100] = 100\n",
    "    z[z < -100] = -100\n",
    "    return 1. / (1 + np.exp(-z))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "class OnlineClassifier(BaseEstimator):\n",
    "    \n",
    "    def __init__(self, C, eta=0.001, n_iter=10):\n",
    "        self.eta = eta\n",
    "        self.n_iter = n_iter\n",
    "        self.loss_ = []\n",
    "        self.weights = []\n",
    "        self.C = C\n",
    "        \n",
    "    def fit(self, X_train, y_train):\n",
    "        a = np.ones((X_train.shape[0],1))\n",
    "        X_train = np.concatenate((a, X_train), axis=1)\n",
    "        w = np.zeros(X_train.shape[1])\n",
    "        for t in range(self.n_iter):\n",
    "            for i in range(X_train.shape[0]):\n",
    "                w_kroneker = [val if idx != 0 else 0 for idx,val in enumerate(w)]\n",
    "                w += self.eta * (self.C * y_train[i] * X_train[i] * float(sigma(-y_train[i]*np.dot(w,X_train[i]))) - w_kroneker)\n",
    "                if (t>0) | (i>0):\n",
    "                    self.weights = np.append(self.weights, np.array(w))\n",
    "                else:\n",
    "                    self.weights = [[w]]\n",
    "                res1 = []\n",
    "                for j in range(X_train.shape[0]):\n",
    "                    res1.append(float(sigma(np.dot(w.T,X_train[j]))))\n",
    "                self.loss_.append(log_loss(y_train,res1))\n",
    "        low_loss_index = self.loss_.index(min(self.loss_))\n",
    "        print(low_loss_index, X_train.shape[1], len(self.weights))\n",
    "        self.w_ = np.array([self.weights[X_train.shape[1]*low_loss_index+i] for i in range(X_train.shape[1])])\n",
    "        return self\n",
    "        \n",
    "    def predict(self, X_test):\n",
    "        prob = self.predict_proba(X_test)\n",
    "        res = []\n",
    "        for i in range(len(prob)):\n",
    "            if prob[i]>= 0.5:\n",
    "                res.append(1)\n",
    "            else:\n",
    "                res.append(-1)\n",
    "        return res\n",
    "        \n",
    "                                 \n",
    "    def predict_proba(self, X_test):\n",
    "        a = np.ones((X_test.shape[0], 1))\n",
    "        X_test = np.concatenate((a, X_test), axis=1)\n",
    "        res1 = []\n",
    "        for j in range(X_test.shape[0]):\n",
    "            res1.append(float(sigma(np.dot(np.array(self.w_),X_test[j]))))\n",
    "        return res1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "16 3 120\n",
      "[ -5.75206787   9.81885873  10.69758197]\n",
      "[0.7390087855654772]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([  5.00045398e+00,   4.95045398e+00,   4.53978687e-04,\n",
       "         5.00045398e+00,   4.95045398e+00,   4.53978687e-04,\n",
       "         5.06732215e+00,   4.90094944e+00,   6.73176137e-02,\n",
       "        -4.87007951e+00,   4.85193994e+00,   6.66444375e-02,\n",
       "         1.75268160e-01,   9.84876821e+00,   6.59779931e-02,\n",
       "         1.75683061e-01,   9.75069543e+00,   6.57331141e-02,\n",
       "         4.57505692e+00,   9.65318848e+00,   4.46444965e+00,\n",
       "        -5.32293719e+00,   9.55665659e+00,   4.41980515e+00,\n",
       "        -5.18002553e+00,   9.60400169e+00,   4.37560710e+00,\n",
       "        -5.17851780e+00,   9.50946941e+00,   4.33335876e+00,\n",
       "         1.81698876e+00,   9.41437471e+00,   1.12855317e+01,\n",
       "        -6.78505534e+00,   9.32023097e+00,   1.11726764e+01,\n",
       "        -6.05076799e+00,   9.96131600e+00,   1.10609496e+01,\n",
       "        -6.05076485e+00,   9.86170599e+00,   1.09503433e+01,\n",
       "        -5.97681850e+00,   9.76308893e+00,   1.09147862e+01,\n",
       "        -6.00212315e+00,   9.66545804e+00,   1.08056384e+01,\n",
       "        -5.75206787e+00,   9.81885873e+00,   1.06975820e+01,\n",
       "        -5.75206400e+00,   9.72067402e+00,   1.05906100e+01,\n",
       "        -5.67350052e+00,   9.62346728e+00,   1.05632674e+01,\n",
       "        -5.70774104e+00,   9.52723261e+00,   1.04576347e+01,\n",
       "        -5.49306136e+00,   9.64663996e+00,   1.03530584e+01,\n",
       "        -5.49305635e+00,   9.55017857e+00,   1.02495328e+01,\n",
       "        -5.40783050e+00,   9.45467678e+00,   1.02322633e+01,\n",
       "        -5.45244409e+00,   9.36013002e+00,   1.01299407e+01,\n",
       "        -5.25553019e+00,   9.46344262e+00,   1.00286413e+01,\n",
       "        -5.25552363e+00,   9.36881476e+00,   9.92836144e+00,\n",
       "        -5.16293186e+00,   9.27512661e+00,   9.92166960e+00,\n",
       "        -5.21985488e+00,   9.18237534e+00,   9.82245290e+00,\n",
       "        -5.03325189e+00,   9.27715458e+00,   9.72422837e+00,\n",
       "        -5.03324331e+00,   9.18439162e+00,   9.62699467e+00,\n",
       "        -4.93310774e+00,   9.09254770e+00,   9.63086030e+00,\n",
       "        -5.00463326e+00,   9.00162223e+00,   9.53455169e+00,\n",
       "        -4.82423856e+00,   9.09200071e+00,   9.43920618e+00,\n",
       "        -4.82422741e+00,   9.00109185e+00,   9.34482526e+00,\n",
       "        -4.71657380e+00,   8.91108093e+00,   9.35903062e+00,\n",
       "        -4.80523839e+00,   8.82197012e+00,   9.26544032e+00,\n",
       "        -4.62830785e+00,   8.91068095e+00,   9.17278591e+00,\n",
       "        -4.62829351e+00,   8.82158848e+00,   9.08107239e+00,\n",
       "        -4.51317264e+00,   8.73337260e+00,   9.10538254e+00,\n",
       "        -4.62161988e+00,   8.64603887e+00,   9.01432871e+00])"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeEAAAFKCAYAAAAqkecjAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fd1d383c898>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "X = np.array([[1., 0.], [1., 1.], [0., 1.], [0., 0.]])\n",
    "y = np.array([1, 1, 1, -1])\n",
    "clf = OnlineClassifier(C=1000, n_iter=10, eta = 0.01)\n",
    "clf.fit(X, y)\n",
    "plt.plot(clf.loss_)\n",
    "print(clf.w_)\n",
    "x1 = np.array([[0.331095, 0.331095]])\n",
    "print(clf.predict_proba(x1))\n",
    "clf.weights"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Проверим `OnlineClassifier` на данных UCI по раку молочной железы."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.datasets import load_breast_cancer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "cancer = load_breast_cancer()\n",
    "# поменяем метки в y с 0 на -1\n",
    "X, y = cancer.data, [-1 if i == 0 else 1 for i in cancer.target]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Выделим 70% под обучение, 30% – под проверку и масштабируем выборку."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "X_train, X_valid, y_train, y_valid = train_test_split(X, y,\n",
    "                                                     test_size=0.3,\n",
    "                                                     random_state=17)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "scaler = StandardScaler()\n",
    "X_train_scaled = scaler.fit_transform(X_train)\n",
    "X_valid_scaled = scaler.transform(X_valid)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Обучите на масштибированной выборке `OnlineClassifier` с параметрами `C`=1, `eta`=$10^{-3}$ и `n_iter`=3."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1193 31 37014\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "OnlineClassifier(C=1, eta=0.001, n_iter=3)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf = OnlineClassifier(C=1, n_iter=3)\n",
    "clf.fit(X_train_scaled, y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Постройте график изменения log_loss."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7fd193c89dd8>]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAecAAAFKCAYAAAAnj5dkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fd196ab2ef0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(clf.loss_)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Теперь обучите `OnlineClassifier` с параметром `C`=1000, число проходов по выборке увеличьте до 10."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3946 31 123380\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "OnlineClassifier(C=1000, eta=0.001, n_iter=10)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "long_clf = OnlineClassifier(C=1000, n_iter=10)\n",
    "long_clf.fit(X_train_scaled, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.048786010364350596"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAdkAAAFKCAYAAABRtSXvAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fd193c2c128>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(long_clf.loss_)\n",
    "min(long_clf.loss_)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Посмотрите на веса модели, при которых ошибка на обучении была минимальна."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font color='red'>Вопрос 3.</font> Какой признак сильнее остальных влияет на вероятность того, что опухоль доброкачественна, согласно обученной модели `OnlineClassifier`? (будьте внимательны – проверьте длину вектора весов, полученного после обучения, сравните с числом признаков в исходной задаче)\n",
    " - worst compactness\n",
    " - worst smoothness\n",
    " - worst concavity\n",
    " - concave points error\n",
    " - concavity error\n",
    " - compactness error\n",
    " - worst fractal dimension"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['mean radius' 'mean texture' 'mean perimeter' 'mean area'\n",
      " 'mean smoothness' 'mean compactness' 'mean concavity'\n",
      " 'mean concave points' 'mean symmetry' 'mean fractal dimension'\n",
      " 'radius error' 'texture error' 'perimeter error' 'area error'\n",
      " 'smoothness error' 'compactness error' 'concavity error'\n",
      " 'concave points error' 'symmetry error' 'fractal dimension error'\n",
      " 'worst radius' 'worst texture' 'worst perimeter' 'worst area'\n",
      " 'worst smoothness' 'worst compactness' 'worst concavity'\n",
      " 'worst concave points' 'worst symmetry' 'worst fractal dimension']\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([ 3.09122261, -1.6575532 , -3.41387355, -1.40982226, -1.66493135,\n",
       "       -0.17720257,  1.97826991, -3.56690964, -3.54327573,  0.4185013 ,\n",
       "        2.86034954, -4.5318188 ,  2.7917743 , -2.57528861, -3.15458401,\n",
       "        0.60303196,  4.53505884,  0.29303788,  0.07966677, -0.52020555,\n",
       "        4.6606677 , -2.92418606, -4.05156371, -1.63955219, -2.51138523,\n",
       "       -1.8587154 , -0.72097361, -5.10734277, -1.97909982, -1.59910147,\n",
       "       -2.29297817])"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(cancer.feature_names)\n",
    "long_clf.w_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Посчитайте log_loss и ROC AUC на отложенной выборке, проделайте все то же с `sklearn.linear_model.LogisticRegression` (параметры по умолчанию, только random_state=17) и сравните результаты."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.434411079396 0.105692134779\n",
      "0.976304023845 0.989269746647\n"
     ]
    }
   ],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "lr = LogisticRegression(random_state = 17)\n",
    "lr.fit(X_train_scaled, y_train)\n",
    "my_pred = long_clf.predict_proba(X_valid_scaled)\n",
    "lr_pred = lr.predict_proba(X_valid_scaled)[:,1]\n",
    "print(log_loss(y_valid,my_pred), log_loss(y_valid, lr_pred))\n",
    "print(roc_auc_score(y_valid,my_pred), roc_auc_score(y_valid, lr_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1.3. Логистическая регрессия и SGDClassifier в задаче классификации отзывов к фильмам"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Теперь посмотрим на логистическую регрессию и ее же версию, но обучаемую стохастическим градиентным спуском, в задаче классификации отзывов IMDB. С этой задачей мы знакомы по 4 и 8 темам курса. Данные можно скачать [отсюда](https://yadi.sk/d/SBnn8s9a3GvYWP).\n",
    "\n",
    "Импортируем файлы, и обучим на имеющихся данных `CountVectorizer`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.datasets import load_files\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "from sklearn.linear_model import SGDClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# поменяйте путь к файлу\n",
    "reviews_train = load_files(\"imdb_reviews/train\")\n",
    "text_train, y_train = reviews_train.data, reviews_train.target"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "reviews_test = load_files(\"imdb_reviews/test\")\n",
    "text_test, y_test = reviews_test.data, reviews_test.target"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Обучим на имеющихся данных `CountVectorizer`, считая биграммы, то есть перейдем к разреженному представлению данных, где каждому уникальному слову и паре подряд идущих слов в обучающей выборке соответсвует признак. Всего таких признаков получается более 1.5 млн."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 39.5 s, sys: 1.21 s, total: 40.7 s\n",
      "Wall time: 43.6 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "cv = CountVectorizer(ngram_range=(1, 2))\n",
    "X_train = cv.fit_transform(text_train)\n",
    "X_test = cv.transform(text_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((25000, 1513832), (25000, 1513832))"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape, X_test.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Обучите на выборке `(X_train, y_train)` логистическую регрессию с параметрами по умолчанию (только укажите `random_state`=17) и посчитайте ROC AUC на тестовой выборке. Замерьте время обучения модели. Данные можно не масштабировать, так как признаки – по сути, счетчики, и они уже все измеряются примерно в одном диапазоне."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 1min 11s, sys: 288 ms, total: 1min 11s\n",
      "Wall time: 1min 12s\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,\n",
       "          intercept_scaling=1, max_iter=100, multi_class='ovr', n_jobs=1,\n",
       "          penalty='l2', random_state=17, solver='liblinear', tol=0.0001,\n",
       "          verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "lr.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.95824855040000001"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred = lr.predict_proba(X_test)[:,1]\n",
    "roc_auc_score(y_test, pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Теперь перейдем к онлайн-алгоритму. Мы написали свой `OnlineClassifier` и принцип его работы поняли, надо еще немного постараться, чтобы сделать его эффективным, например, сделать поддержку разреженных данных. Но мы теперь перейдем к `sklearn`-реализации SGD-алгоритма. Прочитайте документацию [SGDClassifier](http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.SGDClassifier.html), сделайте выводы, чем `SGDClassifier` более продвинут, чем наша реализация онлайн-классификатора. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font color='red'>Вопрос 4.</font> Чем `sklearn`-реализация стохастического классификатора более продвинута, чем `OnlineClassifier`, который мы реализовали? Отметьте все подходящие варианты.\n",
    " - Изменяемый шаг градиентного спуска\n",
    " - Реализован линейный SVM\n",
    " - Реализована ранняя остановка во избежание переобучения\n",
    " - Есть распараллеливание по процессорам\n",
    " - Можно обучать LASSO\n",
    " - Поддерживается онлайн-обучение деревьев решений\n",
    " - Поддерживается mini-batch подход (обновление весов по нескольким объектом сразу, а не по одному)\n"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "1)да - уменьшается\n",
    "2)да\n",
    "3)да\n",
    "4)да\n",
    "5)да\n",
    "6)нет\n",
    "7)да"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Проведите 100 итераций SGD-логрегрессии (опять `random_state`=17) на той же выборке. Опять замерьте время обучения модели и обратите внимание, насколько оно меньше, чем время обучения логистической регрессии."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 10.4 s, sys: 32 ms, total: 10.4 s\n",
      "Wall time: 11.4 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "from sklearn.linear_model import SGDClassifier\n",
    "sgd = SGDClassifier(loss = 'log', penalty = 'l2', n_iter = 100, random_state = 17)\n",
    "sgd.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9582485504\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.95728946560000006"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred1 = sgd.predict_proba(X_test)[:,1]\n",
    "print(roc_auc_score(y_test, pred))\n",
    "roc_auc_score(y_test, pred1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font color='red'>Вопрос 5.</font> В каком знаке после разделителя отличаются ROC AUC на тестовой выборке логистической регрессии и SGD-классификатора с логистической функцией потерь?\n",
    " - 2\n",
    " - 3\n",
    " - 4\n",
    " - 5"
   ]
  }
 ],
 "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}