ML_DD.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Keras usecase"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "from keras.datasets import mnist\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Activation, Flatten\n",
    "from keras.layers import Convolution2D, MaxPooling2D\n",
    "from keras.utils import np_utils"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Возьмем MNIST dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading data from https://s3.amazonaws.com/img-datasets/mnist.pkl.gz\n",
      "15286272/15296311 [============================>.] - ETA: 0s"
     ]
    }
   ],
   "source": [
    "# input image dimensions\n",
    "img_rows, img_cols = 28, 28\n",
    "\n",
    "(X_train, y_train), (X_test, y_test) = mnist.load_data()\n",
    "\n",
    "\n",
    "X_train = X_train.reshape(X_train.shape[0], 1, img_rows, img_cols)\n",
    "X_test = X_test.reshape(X_test.shape[0], 1, img_rows, img_cols)\n",
    "X_train = X_train.astype('float32')\n",
    "X_test = X_test.astype('float32')\n",
    "X_train /= 255\n",
    "X_test /= 255"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x21a8ef850>"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD8CAYAAABXXhlaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x21a587510>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "from matplotlib.pyplot import imshow\n",
    "\n",
    "imshow(X_train[0][0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X_train shape: (60000, 1, 28, 28)\n",
      "60000 train samples\n",
      "10000 test samples\n"
     ]
    }
   ],
   "source": [
    "print 'X_train shape:', X_train.shape \n",
    "print X_train.shape[0], 'train samples'\n",
    "print X_test.shape[0], 'test samples' "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "Y_train = np_utils.to_categorical(y_train, nb_classes)\n",
    "Y_test = np_utils.to_categorical(y_test, nb_classes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Соберем сетку: \n",
    "\n",
    "    1) Для того чтобы она обучалась адекватное время на CPU поставил в начале 2 пулинг слоя, т.е. на вход идут картинки не 28*28, а  всего 7*7. В том числе пожертвовал кол-вом фильтров и кол-вом нейронов в полносвязной сетке.\n",
    "    \n",
    "    2) После сверточного слоя нужно не забывать о ReLU.\n",
    "    \n",
    "    3) После последнего пулинг слоя ставим dropout(отброс 0.25 процентов связей) и flatten слоя, т.е развертку в вектор, чтобы подать на вход полносвязной сетке"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "\n",
    "# number of convolutional filters to use\n",
    "nb_filters = 4\n",
    "# size of pooling area for max pooling\n",
    "nb_pool = 2\n",
    "# convolution kernel size\n",
    "nb_conv = 3\n",
    "\n",
    "\n",
    "batch_size = 128\n",
    "nb_classes = 10\n",
    "nb_epoch = 12\n",
    "\n",
    "model = Sequential()\n",
    "\n",
    "\n",
    "model.add(MaxPooling2D(pool_size=(nb_pool, nb_pool), input_shape=(1, img_rows, img_cols)))\n",
    "model.add(MaxPooling2D(pool_size=(nb_pool, nb_pool)))\n",
    "model.add(Convolution2D(nb_filters, nb_conv, nb_conv,border_mode='valid'))\n",
    "model.add(Activation('relu'))\n",
    "#model.add(Convolution2D(nb_filters, nb_conv, nb_conv))\n",
    "#model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=(nb_pool, nb_pool)))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "model.add(Flatten())\n",
    "model.add(Dense(16))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.5))\n",
    "model.add(Dense(nb_classes))\n",
    "model.add(Activation('softmax'))\n",
    "\n",
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer='adadelta',\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/12\n",
      "60000/60000 [==============================] - 537s - loss: 2.3061 - acc: 0.1201 - val_loss: 2.2257 - val_acc: 0.2229\n",
      "Epoch 2/12\n",
      "60000/60000 [==============================] - 516s - loss: 2.1852 - acc: 0.1901 - val_loss: 2.0303 - val_acc: 0.3302\n",
      "Epoch 3/12\n",
      "60000/60000 [==============================] - 552s - loss: 2.0449 - acc: 0.2566 - val_loss: 1.8246 - val_acc: 0.5015\n",
      "Epoch 4/12\n",
      "60000/60000 [==============================] - 544s - loss: 1.9308 - acc: 0.3047 - val_loss: 1.6819 - val_acc: 0.5430\n",
      "Epoch 5/12\n",
      "60000/60000 [==============================] - 497s - loss: 1.8666 - acc: 0.3307 - val_loss: 1.5960 - val_acc: 0.5659\n",
      "Epoch 6/12\n",
      "60000/60000 [==============================] - 518s - loss: 1.8152 - acc: 0.3477 - val_loss: 1.5290 - val_acc: 0.5820\n",
      "Epoch 7/12\n",
      "60000/60000 [==============================] - 578s - loss: 1.7853 - acc: 0.3588 - val_loss: 1.4855 - val_acc: 0.6017\n",
      "Epoch 8/12\n",
      "60000/60000 [==============================] - 601s - loss: 1.7527 - acc: 0.3704 - val_loss: 1.4328 - val_acc: 0.6180\n",
      "Epoch 9/12\n",
      "60000/60000 [==============================] - 541s - loss: 1.7206 - acc: 0.3836 - val_loss: 1.3857 - val_acc: 0.6331\n",
      "Epoch 10/12\n",
      "60000/60000 [==============================] - 539s - loss: 1.6944 - acc: 0.3931 - val_loss: 1.3555 - val_acc: 0.6504\n",
      "Epoch 11/12\n",
      "60000/60000 [==============================] - 511s - loss: 1.6730 - acc: 0.4025 - val_loss: 1.3201 - val_acc: 0.6574\n",
      "Epoch 12/12\n",
      "60000/60000 [==============================] - 509s - loss: 1.6537 - acc: 0.4096 - val_loss: 1.2876 - val_acc: 0.6637\n",
      "('Test score:', 1.2875659545898437)\n",
      "('Test accuracy:', 0.66369999999999996)\n"
     ]
    }
   ],
   "source": [
    "model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch,\n",
    "          verbose=1, validation_data=(X_test, Y_test))\n",
    "score = model.evaluate(X_test, Y_test, verbose=0)\n",
    "print('Test score:', score[0])\n",
    "print('Test accuracy:', score[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##  Word2vec"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import gensim\n",
    "\n",
    "model = gensim.models.Word2Vec.load_word2vec_format('./Downloads/GoogleNews-vectors-negative300.bin', binary=True)  \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    " Поиск синонимов"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(u'Credit', 0.689106822013855),\n",
       " (u'loan', 0.5559651851654053),\n",
       " (u'loans', 0.5492815971374512),\n",
       " (u'lending', 0.5242205262184143),\n",
       " (u'mortgage', 0.5069109201431274),\n",
       " (u'borrowers', 0.4965798854827881),\n",
       " (u'lenders', 0.49586641788482666),\n",
       " (u'liquidity', 0.4858595132827759),\n",
       " (u'debt', 0.48311686515808105),\n",
       " (u'CREDIT', 0.4826412498950958)]"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.most_similar('credit')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(u'Government', 0.7132059335708618),\n",
       " (u'goverment', 0.7049153447151184),\n",
       " (u'governent', 0.6665106415748596),\n",
       " (u'governments', 0.6521533727645874),\n",
       " (u'govern_ment', 0.6326169967651367),\n",
       " (u'governmnent', 0.6278249621391296),\n",
       " (u'governement', 0.6268108487129211),\n",
       " (u'govenrment', 0.6198933124542236),\n",
       " (u'governemnt', 0.6029759049415588),\n",
       " (u'govt', 0.597993791103363)]"
      ]
     },
     "execution_count": 59,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.most_similar('government')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Арифметические операции"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(u'queen', 0.7118192911148071),\n",
       " (u'monarch', 0.6189674139022827),\n",
       " (u'princess', 0.5902431607246399),\n",
       " (u'crown_prince', 0.5499460697174072),\n",
       " (u'prince', 0.5377321243286133),\n",
       " (u'kings', 0.5236844420433044),\n",
       " (u'Queen_Consort', 0.5235945582389832),\n",
       " (u'queens', 0.5181134343147278),\n",
       " (u'sultan', 0.5098593235015869),\n",
       " (u'monarchy', 0.5087411999702454)]"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.most_similar(positive=['woman', 'king'], negative=['man'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(u'France', 0.7609949111938477),\n",
       " (u'Belgium', 0.5987190008163452),\n",
       " (u'Villebon_Sur_Yvette', 0.5668909549713135),\n",
       " (u'extradites_Noriega', 0.5622164011001587),\n",
       " (u'French', 0.5535142421722412),\n",
       " (u'Italy', 0.5476635694503784),\n",
       " (u'called_Xynthia_blew', 0.5446068048477173),\n",
       " (u'Brive_la', 0.5406582951545715),\n",
       " (u'Morocco', 0.5371377468109131),\n",
       " (u'Dordogne_region', 0.5348877906799316)]"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.most_similar(positive=['Spain', 'Paris'], negative=['Madrid'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(u'Russia', 0.5521125793457031),\n",
       " (u'Ukraine', 0.5152689814567566),\n",
       " (u'lifts_Squaw_Valley', 0.48065853118896484),\n",
       " (u'Byelorussia', 0.45393040776252747),\n",
       " (u'Evgeni_Popov', 0.44755619764328003),\n",
       " (u'lifts_Sugarloaf', 0.44729477167129517),\n",
       " (u'Grigory_Falko', 0.4464748501777649),\n",
       " (u'Russian', 0.4459983706474304),\n",
       " (u'Kazakhstan', 0.44056546688079834),\n",
       " (u'proudly_proclaims_Made', 0.4373660683631897)]"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.most_similar(positive=['USA', 'Putin'], negative=['Obama'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}