{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<center><img src=\"img/logo_hse_black.jpg\"></center>\n",
    "\n",
    "<h1><center>Методы машинного обучения</center></h1>\n",
    "<h2><center>Семинар: линейные модели классификации</center></h2>\n",
    "<h2><center>Методы понижения размерности</center></h2>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.style.use('ggplot')\n",
    "plt.rcParams['figure.figsize'] = (12,8)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Логистическая регрессия"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Игрушечный пример"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Сгенерируем выборку и опробуем логистическую регрессию"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(0)\n",
    "X = np.r_[np.random.randn(20, 2) + [2, 2],\n",
    "          np.random.randn(20, 2) + [-2, -2]]\n",
    "y = [-1] * 20 + [1] * 20"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x115124f98>"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 504x504 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(7, 7))\n",
    "ax.scatter(X[:, 0],\n",
    "           X[:, 1],\n",
    "           c=y,\n",
    "           cmap=plt.cm.Paired)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LogisticRegression"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Обучим логистическую регрессию на этих данных и нарисуем разделяющую гиперплоскость"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\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='warn',\n",
       "          n_jobs=None, penalty='l2', random_state=None, solver='warn',\n",
       "          tol=0.0001, verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = LogisticRegression(C=1.0, \n",
    "                           fit_intercept=True, \n",
    "                           penalty='l2')\n",
    "model.fit(X, y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "w_0 = -0.183954\n",
      "w_1, w_2 =  [[-1.06097157 -1.00171289]]\n"
     ]
    }
   ],
   "source": [
    "print('w_0 = %f' % model.intercept_)\n",
    "print('w_1, w_2 = ', model.coef_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Нарисуем эту гиперплоскость\n",
    "w_0 = model.intercept_[0]\n",
    "w_1 = model.coef_[0][0]\n",
    "w_2 = model.coef_[0][1]\n",
    "\n",
    "x_1 = np.linspace(-4, 4, 10)\n",
    "x_2 = - (w_0 + w_1*x_1)/w_2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1a17b1f8d0>]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 504x504 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(7, 7))\n",
    "ax.scatter(X[:, 0],\n",
    "           X[:, 1],\n",
    "           c=y,\n",
    "           cmap=plt.cm.Paired)\n",
    "plt.plot(x_1, x_2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Пример с текстами"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Возьмем текстовые данные [отсюда](https://archive.ics.uci.edu/ml/machine-learning-databases/00331/). Архив содержит 3 файла с положительными и отрицательными отзывами с ресурсов\n",
    "* imdb.com\n",
    "* amazon.com\n",
    "* yelp.com\n",
    "\n",
    "Формат файла следующий:\n",
    "<отзыв>\\t<метка>\\n\n",
    "\n",
    "\n",
    "### Задача\n",
    "1. Загрузите тексты и метки классов в разные переменные\n",
    "2. Выберите меру качества классификации\n",
    "3. Обучите логистическую (без подбора гиперпараметров). Тексты представляются в виде мешка слов\n",
    "4. Выведите наиболее значимые слова из текста\n",
    "5. С помощью кросс-валидации найдите хорошие значения гиперпараметров для `CountVectorizer` и `LogisticRegression`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv('data/sentiment/imdb_labelled.txt', sep='\\t', header=None, names=['text', 'label'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "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>text</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>A very, very, very slow-moving, aimless movie ...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Not sure who was more lost - the flat characte...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Attempting artiness with black &amp; white and cle...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Very little music or anything to speak of.</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>The best scene in the movie was when Gerardo i...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                text  label\n",
       "0  A very, very, very slow-moving, aimless movie ...      0\n",
       "1  Not sure who was more lost - the flat characte...      0\n",
       "2  Attempting artiness with black & white and cle...      0\n",
       "3       Very little music or anything to speak of.        0\n",
       "4  The best scene in the movie was when Gerardo i...      1"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.model_selection import StratifiedKFold\n",
    "from sklearn.model_selection import RandomizedSearchCV, GridSearchCV"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Pipeline([\n",
    "    ('vect', CountVectorizer(max_df=0.95, stop_words='english', ngram_range=(2,2))),\n",
    "    ('clf', LogisticRegression())\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = df.text.values\n",
    "y = df.label.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Pipeline(memory=None,\n",
       "     steps=[('vect', CountVectorizer(analyzer='word', binary=False, decode_error='strict',\n",
       "        dtype=<class 'numpy.int64'>, encoding='utf-8', input='content',\n",
       "        lowercase=True, max_df=0.95, max_features=None, min_df=1,\n",
       "        ngram_range=(2, 2), preprocessor=None, stop_words='english',\n",
       "       ...penalty='l2', random_state=None, solver='warn',\n",
       "          tol=0.0001, verbose=0, warm_start=False))])"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(X, y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "coefs = model.named_steps['clf'].coef_[0] # Веса линейной регрессии для каждого словa\n",
    "words = model.named_steps['vect'].get_feature_names()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "word_coefs = pd.Series(index=words, data=coefs).sort_values()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "waste time               -0.945444\n",
       "bad film                 -0.801509\n",
       "avoid costs              -0.655693\n",
       "just bad                 -0.621132\n",
       "don waste                -0.585915\n",
       "film just                -0.565970\n",
       "worst series             -0.554764\n",
       "90 minutes               -0.549202\n",
       "hour half                -0.545175\n",
       "cover girl               -0.521187\n",
       "story stupid             -0.513814\n",
       "avoid avoid              -0.512680\n",
       "movie pretty             -0.488534\n",
       "ve seen                  -0.462487\n",
       "just awful               -0.456271\n",
       "acting bad               -0.447700\n",
       "film started             -0.444112\n",
       "really bad               -0.424706\n",
       "hate movies              -0.418471\n",
       "just didn                -0.414149\n",
       "action scenes            -0.413278\n",
       "couldn seriously         -0.393361\n",
       "worse ticker             -0.393361\n",
       "just lame                -0.393361\n",
       "plot storyline           -0.393361\n",
       "just painful             -0.393361\n",
       "great disappointment     -0.393361\n",
       "script script            -0.393361\n",
       "plot whatsoever          -0.393361\n",
       "engaging exciting        -0.393361\n",
       "                            ...   \n",
       "transfers good            0.408732\n",
       "like movie                0.412289\n",
       "rate movie                0.417140\n",
       "movie 10                  0.417140\n",
       "film like                 0.417562\n",
       "think disappointed        0.420384\n",
       "characters interesting    0.423554\n",
       "wonderful story           0.436351\n",
       "feel good                 0.455643\n",
       "cinematography film       0.463604\n",
       "excellent job             0.487130\n",
       "thought provoking         0.487886\n",
       "excellent film            0.491852\n",
       "wind lion                 0.496507\n",
       "short film                0.496581\n",
       "cast good                 0.500491\n",
       "movie really              0.507134\n",
       "great film                0.519239\n",
       "music film                0.526220\n",
       "liked movie               0.540232\n",
       "definitely worth          0.559673\n",
       "highly recommended        0.568294\n",
       "saw film                  0.573340\n",
       "movie good                0.574355\n",
       "long time                 0.598076\n",
       "excellent performance     0.643658\n",
       "gave 10                   0.643658\n",
       "worth checking            0.761545\n",
       "film great                0.837645\n",
       "10 10                     1.126112\n",
       "Length: 5693, dtype: float64"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "word_coefs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Методы понижения размерности"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load data\n",
    "df_wine = pd.read_csv('data/winequality-red.csv', sep=';')\n",
    "\n",
    "# Make classification target feature\n",
    "df_wine.loc[:, 'quality_cat'] = (df_wine.quality > 5).astype(int)\n",
    "df_wine = df_wine.drop('quality', axis=1)\n",
    "\n",
    "# Get descriptive and target features\n",
    "X = df_wine.iloc[:, :-1].values\n",
    "y = df_wine.iloc[:, -1].values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "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>fixed acidity</th>\n",
       "      <th>volatile acidity</th>\n",
       "      <th>citric acid</th>\n",
       "      <th>residual sugar</th>\n",
       "      <th>chlorides</th>\n",
       "      <th>free sulfur dioxide</th>\n",
       "      <th>total sulfur dioxide</th>\n",
       "      <th>density</th>\n",
       "      <th>pH</th>\n",
       "      <th>sulphates</th>\n",
       "      <th>alcohol</th>\n",
       "      <th>quality_cat</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>7.4</td>\n",
       "      <td>0.70</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.9</td>\n",
       "      <td>0.076</td>\n",
       "      <td>11.0</td>\n",
       "      <td>34.0</td>\n",
       "      <td>0.9978</td>\n",
       "      <td>3.51</td>\n",
       "      <td>0.56</td>\n",
       "      <td>9.4</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>7.8</td>\n",
       "      <td>0.88</td>\n",
       "      <td>0.00</td>\n",
       "      <td>2.6</td>\n",
       "      <td>0.098</td>\n",
       "      <td>25.0</td>\n",
       "      <td>67.0</td>\n",
       "      <td>0.9968</td>\n",
       "      <td>3.20</td>\n",
       "      <td>0.68</td>\n",
       "      <td>9.8</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>7.8</td>\n",
       "      <td>0.76</td>\n",
       "      <td>0.04</td>\n",
       "      <td>2.3</td>\n",
       "      <td>0.092</td>\n",
       "      <td>15.0</td>\n",
       "      <td>54.0</td>\n",
       "      <td>0.9970</td>\n",
       "      <td>3.26</td>\n",
       "      <td>0.65</td>\n",
       "      <td>9.8</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>11.2</td>\n",
       "      <td>0.28</td>\n",
       "      <td>0.56</td>\n",
       "      <td>1.9</td>\n",
       "      <td>0.075</td>\n",
       "      <td>17.0</td>\n",
       "      <td>60.0</td>\n",
       "      <td>0.9980</td>\n",
       "      <td>3.16</td>\n",
       "      <td>0.58</td>\n",
       "      <td>9.8</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>7.4</td>\n",
       "      <td>0.70</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.9</td>\n",
       "      <td>0.076</td>\n",
       "      <td>11.0</td>\n",
       "      <td>34.0</td>\n",
       "      <td>0.9978</td>\n",
       "      <td>3.51</td>\n",
       "      <td>0.56</td>\n",
       "      <td>9.4</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   fixed acidity  volatile acidity  citric acid  residual sugar  chlorides  \\\n",
       "0            7.4              0.70         0.00             1.9      0.076   \n",
       "1            7.8              0.88         0.00             2.6      0.098   \n",
       "2            7.8              0.76         0.04             2.3      0.092   \n",
       "3           11.2              0.28         0.56             1.9      0.075   \n",
       "4            7.4              0.70         0.00             1.9      0.076   \n",
       "\n",
       "   free sulfur dioxide  total sulfur dioxide  density    pH  sulphates  \\\n",
       "0                 11.0                  34.0   0.9978  3.51       0.56   \n",
       "1                 25.0                  67.0   0.9968  3.20       0.68   \n",
       "2                 15.0                  54.0   0.9970  3.26       0.65   \n",
       "3                 17.0                  60.0   0.9980  3.16       0.58   \n",
       "4                 11.0                  34.0   0.9978  3.51       0.56   \n",
       "\n",
       "   alcohol  quality_cat  \n",
       "0      9.4            0  \n",
       "1      9.8            0  \n",
       "2      9.8            0  \n",
       "3      9.8            1  \n",
       "4      9.4            0  "
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_wine.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Recursive Feature Elimination"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.feature_selection import RFECV\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.model_selection import StratifiedKFold\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.preprocessing import StandardScaler"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_rfe = Pipeline([\n",
    "    ('scaler', StandardScaler()),\n",
    "    ('clf', RFECV(LogisticRegression(), step=1, cv=cv, verbose=1, scoring='roc_auc'))\n",
    "])\n",
    "\n",
    "cv = StratifiedKFold(n_splits=5, shuffle=True, random_state=123)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting estimator with 11 features.\n",
      "Fitting estimator with 10 features.\n",
      "Fitting estimator with 9 features.\n",
      "Fitting estimator with 8 features.\n",
      "Fitting estimator with 7 features.\n",
      "Fitting estimator with 6 features.\n",
      "Fitting estimator with 5 features.\n",
      "Fitting estimator with 4 features.\n",
      "Fitting estimator with 3 features.\n",
      "Fitting estimator with 2 features.\n",
      "Fitting estimator with 11 features.\n",
      "Fitting estimator with 10 features.\n",
      "Fitting estimator with 9 features.\n",
      "Fitting estimator with 8 features.\n",
      "Fitting estimator with 7 features.\n",
      "Fitting estimator with 6 features.\n",
      "Fitting estimator with 5 features.\n",
      "Fitting estimator with 4 features.\n",
      "Fitting estimator with 3 features.\n",
      "Fitting estimator with 2 features.\n",
      "Fitting estimator with 11 features.\n",
      "Fitting estimator with 10 features.\n",
      "Fitting estimator with 9 features.\n",
      "Fitting estimator with 8 features.\n",
      "Fitting estimator with 7 features.\n",
      "Fitting estimator with 6 features.\n",
      "Fitting estimator with 5 features.\n",
      "Fitting estimator with 4 features.\n",
      "Fitting estimator with 3 features.\n",
      "Fitting estimator with 2 features.\n",
      "Fitting estimator with 11 features.\n",
      "Fitting estimator with 10 features.\n",
      "Fitting estimator with 9 features.\n",
      "Fitting estimator with 8 features.\n",
      "Fitting estimator with 7 features.\n",
      "Fitting estimator with 6 features.\n",
      "Fitting estimator with 5 features.\n",
      "Fitting estimator with 4 features.\n",
      "Fitting estimator with 3 features.\n",
      "Fitting estimator with 2 features.\n",
      "Fitting estimator with 11 features.\n",
      "Fitting estimator with 10 features.\n",
      "Fitting estimator with 9 features.\n",
      "Fitting estimator with 8 features.\n",
      "Fitting estimator with 7 features.\n",
      "Fitting estimator with 6 features.\n",
      "Fitting estimator with 5 features.\n",
      "Fitting estimator with 4 features.\n",
      "Fitting estimator with 3 features.\n",
      "Fitting estimator with 2 features.\n",
      "Fitting estimator with 11 features.\n",
      "Fitting estimator with 10 features.\n",
      "Fitting estimator with 9 features.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n",
      "/Users/andrey.shestakov/anaconda3/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:432: FutureWarning: Default solver will be changed to 'lbfgs' in 0.22. Specify a solver to silence this warning.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Pipeline(memory=None,\n",
       "     steps=[('scaler', StandardScaler(copy=True, with_mean=True, with_std=True)), ('clf', RFECV(cv=StratifiedKFold(n_splits=5, random_state=123, shuffle=True),\n",
       "   estimator=LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,\n",
       "          intercept_scaling=1, max_iter=100, multi_clas...m_start=False),\n",
       "   min_features_to_select=1, n_jobs=None, scoring='roc_auc', step=1,\n",
       "   verbose=1))])"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_rfe.fit(X, y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "rfe = model_rfe.named_steps['clf']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.75720164, 0.7955729 , 0.80091782, 0.81279034, 0.81248978,\n",
       "       0.81463012, 0.8164726 , 0.81848756, 0.81790802, 0.81704451,\n",
       "       0.81717928])"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rfe.grid_scores_ # Качество для 1, 2, 3 и тп признаков"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ True,  True,  True, False,  True,  True,  True, False, False,\n",
       "        True,  True])"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rfe.support_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_of_features = range(1, 12)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0,0.5,'ROC AUC')"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(num_of_features, rfe.grid_scores_ )\n",
    "plt.xlabel('Num of features')\n",
    "plt.ylabel('ROC AUC')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## PCA"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Попробуем получить PCA разными способами"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## PCA через sklearn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.decomposition import PCA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pca = PCA(n_components=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## PCA через ковариационную матрицу"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from numpy.linalg import eig"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Оценка качества при разных количествах компонент\n",
    "\n",
    "Реализуйте 2 пайплайнк:\n",
    "    * StandartScaler + LogisticRegression\n",
    "    * StandartScaler + PCA + LogisticRegression\n",
    "\n",
    "Оцените качество пайплайна с PCA при разных количествах компонент и сравните его с первым"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.model_selection import StratifiedKFold, cross_val_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Singular Value Decomposition\n",
    "\n",
    "Для любой матрицы $A$ размера $n \\times m$ и ранга $r$ можно найти разложение вида:\n",
    "$$ A = U \\Sigma V^\\top ,$$\n",
    "где \n",
    "* $U$ - унитарная матрица, состоящая из собственных векторов $AA^\\top$\n",
    "* $V$ - унитарная матрица, состоящая из собственных векторов $A^\\top A$\n",
    "* $\\Sigma$ - диагональная матрица с сингулярными числами $s_i = \\sqrt{\\lambda_i}$\n",
    "\n",
    "<center><img src='img/svd.jpg' width=600></center>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## SVD via PCA\n",
    "Матрицы $U$ и $V$ ортогональны и могут быть использованы для перехода к ортогональному базису:\n",
    "$$ AV = U\\Sigma $$\n",
    "\n",
    "<center><img src='img/pca_svd.png' width=600></center>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from numpy.linalg import svd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Рационы питания в странах"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Загрузите набор данных о пищевом рационе в разных странах мира `diet.csv`\n",
    "* Примените на данных PCA с 2 компонентами\n",
    "* Изобразите объекты в сжатом пространстве"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "df = pd.read_csv('data/diet.csv', sep=';')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Скорее всего вы обнаружите некоторые выбросы, с этим ничего не поделать - PCA чувствителен к выбросам\n",
    "* Удалите объекты-выборосы и повторите процедуру"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# PCA для текстов\n",
    "#### Он же Latent Semantic Index или Latent Semantic Allocation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Когда мы работаем с текстами, бы обычно переходим к модели мешка слов\n",
    "* Что если мы ходим применить для текстов PCA?\n",
    "* Центрировать данные мы не можем, так как тогда наша огромная матрица с признаками будет плотной, а значит будет очень много весить\n",
    "* Давайте не будем центрировать! =D"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.decomposition import TruncatedSVD"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# T-SNE"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Задание\n",
    "Сожмите признаковое пространство данных с цифрами с помощью tsne."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.datasets import load_digits"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "## Your Code Here"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "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": "3.7.0"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {
    "height": "142px",
    "width": "252px"
   },
   "number_sections": false,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": "block",
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}