Movie reviews classification with neural networks.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "from keras.datasets import imdb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "(train_data, train_labels), (test_data, test_labels) = imdb.load_data(\n",
    "num_words=10000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def vectorize_sequences(sequences, dimension=10000):\n",
    "    # Creates an all-zero matrix of shape (len(sequences),dimension)\n",
    "    results = np.zeros((len(sequences), dimension))\n",
    "    for i, sequence in enumerate(sequences):\n",
    "       #Sets specific indicesof results[i] to 1s\n",
    "        results[i, sequence] = 1.\n",
    "    return results\n",
    "\n",
    "#Vectorized training dat\n",
    "x_train = vectorize_sequences(train_data)\n",
    "x_test = vectorize_sequences(test_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0., 1., 1., ..., 0., 0., 0.],\n",
       "       [0., 1., 1., ..., 0., 0., 0.],\n",
       "       [0., 1., 1., ..., 0., 0., 0.],\n",
       "       [0., 1., 1., ..., 0., 0., 0.],\n",
       "       [0., 1., 1., ..., 0., 0., 0.]])"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_train[0:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0, 1, 1, 0, 1])"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_labels[0:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train = np.asarray(train_labels).astype('float32')\n",
    "y_test = np.asarray(test_labels).astype('float32')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0, 1, 1, 0, 1])"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_labels[0:5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### The input data is now vectors, and the labels are scalars (1s and 0s)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras import models\n",
    "from keras import layers\n",
    "from keras import optimizers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = models.Sequential()\n",
    "model.add(layers.Dense(16, activation='relu', input_shape=(10000,)))\n",
    "model.add(layers.Dense(16, activation='relu'))\n",
    "model.add(layers.Dense(1, activation='sigmoid'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras import losses\n",
    "from keras import metrics"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "crossentropy is usually the best choice when you’re dealing\n",
    "with models that output probabilities. Crossentropy is a quantity from the field of Information Theory that measures the distance between probability distributions or, in this\n",
    "case, between the ground-truth distribution and your predictions.   "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "opt = tf.keras.optimizers.RMSprop(lr=0.001)\n",
    "\n",
    "model.compile(optimizer=opt,\n",
    "    loss=losses.binary_crossentropy,\n",
    "    metrics=[metrics.binary_accuracy])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "#taking first 10 000 sample as valdiation data fron x and y\n",
    "x_val = x_train[:10000]\n",
    "partial_x_train = x_train[10000:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_val = y_train[:10000]\n",
    "partial_y_train = y_train[10000:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "30/30 [==============================] - 2s 52ms/step - loss: 0.5193 - acc: 0.7736 - val_loss: 0.3841 - val_acc: 0.8708\n",
      "Epoch 2/20\n",
      "30/30 [==============================] - 1s 17ms/step - loss: 0.3094 - acc: 0.9050 - val_loss: 0.3244 - val_acc: 0.8762\n",
      "Epoch 3/20\n",
      "30/30 [==============================] - 1s 20ms/step - loss: 0.2272 - acc: 0.9289 - val_loss: 0.2838 - val_acc: 0.8912\n",
      "Epoch 4/20\n",
      "30/30 [==============================] - 0s 17ms/step - loss: 0.1756 - acc: 0.9443 - val_loss: 0.2764 - val_acc: 0.8907\n",
      "Epoch 5/20\n",
      "30/30 [==============================] - 1s 20ms/step - loss: 0.1447 - acc: 0.9549 - val_loss: 0.2827 - val_acc: 0.8870\n",
      "Epoch 6/20\n",
      "30/30 [==============================] - 1s 17ms/step - loss: 0.1192 - acc: 0.9637 - val_loss: 0.2905 - val_acc: 0.8871\n",
      "Epoch 7/20\n",
      "30/30 [==============================] - 1s 18ms/step - loss: 0.1009 - acc: 0.9716 - val_loss: 0.3091 - val_acc: 0.8824\n",
      "Epoch 8/20\n",
      "30/30 [==============================] - 0s 16ms/step - loss: 0.0809 - acc: 0.9785 - val_loss: 0.3256 - val_acc: 0.8821\n",
      "Epoch 9/20\n",
      "30/30 [==============================] - 0s 16ms/step - loss: 0.0682 - acc: 0.9833 - val_loss: 0.3501 - val_acc: 0.8769\n",
      "Epoch 10/20\n",
      "30/30 [==============================] - 1s 17ms/step - loss: 0.0568 - acc: 0.9869 - val_loss: 0.3847 - val_acc: 0.8756\n",
      "Epoch 11/20\n",
      "30/30 [==============================] - 1s 18ms/step - loss: 0.0490 - acc: 0.9876 - val_loss: 0.3919 - val_acc: 0.8757\n",
      "Epoch 12/20\n",
      "30/30 [==============================] - 1s 17ms/step - loss: 0.0374 - acc: 0.9923 - val_loss: 0.4657 - val_acc: 0.8635\n",
      "Epoch 13/20\n",
      "30/30 [==============================] - 1s 18ms/step - loss: 0.0333 - acc: 0.9933 - val_loss: 0.4551 - val_acc: 0.8707\n",
      "Epoch 14/20\n",
      "30/30 [==============================] - 0s 16ms/step - loss: 0.0275 - acc: 0.9944 - val_loss: 0.4704 - val_acc: 0.8708\n",
      "Epoch 15/20\n",
      "30/30 [==============================] - 0s 17ms/step - loss: 0.0205 - acc: 0.9971 - val_loss: 0.4960 - val_acc: 0.8716\n",
      "Epoch 16/20\n",
      "30/30 [==============================] - 0s 16ms/step - loss: 0.0168 - acc: 0.9977 - val_loss: 0.5260 - val_acc: 0.8715\n",
      "Epoch 17/20\n",
      "30/30 [==============================] - 0s 16ms/step - loss: 0.0137 - acc: 0.9982 - val_loss: 0.5573 - val_acc: 0.8689\n",
      "Epoch 18/20\n",
      "30/30 [==============================] - 0s 16ms/step - loss: 0.0136 - acc: 0.9979 - val_loss: 0.5973 - val_acc: 0.8656\n",
      "Epoch 19/20\n",
      "30/30 [==============================] - 1s 17ms/step - loss: 0.0066 - acc: 0.9997 - val_loss: 0.6281 - val_acc: 0.8642\n",
      "Epoch 20/20\n",
      "30/30 [==============================] - 1s 17ms/step - loss: 0.0078 - acc: 0.9994 - val_loss: 0.6524 - val_acc: 0.8663\n"
     ]
    }
   ],
   "source": [
    "model.compile(optimizer='rmsprop',\n",
    "    loss='binary_crossentropy',\n",
    "    metrics=['acc'])\n",
    "history = model.fit(partial_x_train,\n",
    "    partial_y_train,\n",
    "    epochs=20,\n",
    "    batch_size=512,\n",
    "    validation_data=(x_val, y_val))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "the history objectis a dictionary containing data about everything that happened during training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    " history_dict = history.history"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_keys(['loss', 'acc', 'val_loss', 'val_acc'])"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    " history_dict.keys()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "history_dict = history.history\n",
    "loss_values = history_dict['loss']\n",
    "val_loss_values = history_dict['val_loss']\n",
    "acc = history_dict['acc']\n",
    "\n",
    "epochs = range(1, len(acc) + 1)\n",
    "\n",
    "plt.plot(epochs, loss_values, 'go', label='Training loss')\n",
    "plt.plot(epochs, val_loss_values, 'b', label='Validation loss')\n",
    "\n",
    "plt.title('Training and validation loss')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.clf()\n",
    "\n",
    "acc_values = history_dict['acc']\n",
    "val_acc = history_dict['val_acc']\n",
    "plt.plot(epochs, acc, 'go', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Retraining a model from scratch"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "after the second epoch, we're overoptimizing on the training data, and we end up learning representations that are specific to the training data and don’t generalize to data outside of the training set.\n",
    "In this case, to prevent overfitting, we could stop training after three epochs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/4\n",
      "49/49 [==============================] - 1s 13ms/step - loss: 0.4580 - accuracy: 0.8259\n",
      "Epoch 2/4\n",
      "49/49 [==============================] - 1s 11ms/step - loss: 0.2625 - accuracy: 0.9116\n",
      "Epoch 3/4\n",
      "49/49 [==============================] - 1s 11ms/step - loss: 0.2007 - accuracy: 0.9302\n",
      "Epoch 4/4\n",
      "49/49 [==============================] - 1s 11ms/step - loss: 0.1692 - accuracy: 0.9401\n",
      "782/782 [==============================] - 1s 1ms/step - loss: 0.2943 - accuracy: 0.8832\n"
     ]
    }
   ],
   "source": [
    "model = models.Sequential()\n",
    "model.add(layers.Dense(16, activation='relu', input_shape=(10000,)))\n",
    "model.add(layers.Dense(16, activation='relu'))\n",
    "model.add(layers.Dense(1, activation='sigmoid'))\n",
    "\n",
    "model.compile(optimizer='rmsprop',\n",
    "    loss='binary_crossentropy',\n",
    "    metrics=['accuracy'])\n",
    "\n",
    "model.fit(x_train, y_train, epochs=4, batch_size=512)\n",
    "results = model.evaluate(x_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.29432356357574463, 0.8832399845123291]"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.19722128],\n",
       "       [0.9989947 ],\n",
       "       [0.9361497 ],\n",
       "       ...,\n",
       "       [0.13622987],\n",
       "       [0.0975121 ],\n",
       "       [0.7170367 ]], dtype=float32)"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.predict(x_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}