MLP for Text Classification.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import keras"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Processing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "8982 train sequences\n",
      "2246 test sequences\n"
     ]
    }
   ],
   "source": [
    "from keras.datasets import reuters\n",
    "\n",
    "(input_train, y_train), (input_test, y_test) = reuters.load_data(num_words=None, test_split=0.2)\n",
    "print(len(input_train), 'train sequences')\n",
    "print(len(input_test), 'test sequences')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "input_train shape: (8982, 500)\n",
      "input_test shape: (2246, 500)\n",
      "y_train shape: (8982,)\n",
      "y_test shape: (2246,)\n"
     ]
    }
   ],
   "source": [
    "from keras.preprocessing import sequence\n",
    "\n",
    "maxlen = 500  # cut texts after this number of words\n",
    "\n",
    "input_train = sequence.pad_sequences(input_train, maxlen=maxlen)\n",
    "input_test = sequence.pad_sequences(input_test, maxlen=maxlen)\n",
    "\n",
    "print('input_train shape:', input_train.shape)\n",
    "print('input_test shape:', input_test.shape)\n",
    "print('y_train shape:', y_train.shape)\n",
    "print('y_test shape:', y_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# of Classes:  46\n"
     ]
    }
   ],
   "source": [
    "num_classes = max(y_train) + 1\n",
    "print('# of Classes: ',num_classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.utils import np_utils\n",
    "\n",
    "y_train = np_utils.to_categorical(y_train, num_classes)\n",
    "y_test = np_utils.to_categorical(y_test, num_classes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Multi-Layer Perceptron (MLP) Model Setup"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /home/pedrobalage/Documents/portoai-nlp/venv/lib/python3.6/site-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Colocations handled automatically by placer.\n",
      "WARNING:tensorflow:From /home/pedrobalage/Documents/portoai-nlp/venv/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Activation\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Dense(16, input_shape=(maxlen,)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.5))\n",
    "model.add(Dense(num_classes))\n",
    "model.add(Activation('softmax'))\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_1 (Dense)              (None, 16)                8016      \n",
      "_________________________________________________________________\n",
      "activation_1 (Activation)    (None, 16)                0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 16)                0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 46)                782       \n",
      "_________________________________________________________________\n",
      "activation_2 (Activation)    (None, 46)                0         \n",
      "=================================================================\n",
      "Total params: 8,798\n",
      "Trainable params: 8,798\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /home/pedrobalage/Documents/portoai-nlp/venv/lib/python3.6/site-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.cast instead.\n",
      "Train on 6287 samples, validate on 2695 samples\n",
      "Epoch 1/20\n",
      "6287/6287 [==============================] - 0s 44us/step - loss: 14.5110 - acc: 0.0907 - val_loss: 12.4006 - val_acc: 0.2249\n",
      "Epoch 2/20\n",
      "6287/6287 [==============================] - 0s 19us/step - loss: 12.6783 - acc: 0.2028 - val_loss: 11.0266 - val_acc: 0.3147\n",
      "Epoch 3/20\n",
      "6287/6287 [==============================] - 0s 18us/step - loss: 11.7609 - acc: 0.2574 - val_loss: 10.6417 - val_acc: 0.3395\n",
      "Epoch 4/20\n",
      "6287/6287 [==============================] - 0s 19us/step - loss: 11.0127 - acc: 0.2973 - val_loss: 10.5740 - val_acc: 0.3440\n",
      "Epoch 5/20\n",
      "6287/6287 [==============================] - 0s 18us/step - loss: 10.6946 - acc: 0.3086 - val_loss: 10.5017 - val_acc: 0.3484\n",
      "Epoch 6/20\n",
      "6287/6287 [==============================] - 0s 18us/step - loss: 10.4843 - acc: 0.3242 - val_loss: 10.4842 - val_acc: 0.3495\n",
      "Epoch 7/20\n",
      "6287/6287 [==============================] - 0s 18us/step - loss: 10.4212 - acc: 0.3230 - val_loss: 10.4723 - val_acc: 0.3503\n",
      "Epoch 8/20\n",
      "6287/6287 [==============================] - 0s 18us/step - loss: 10.2871 - acc: 0.3329 - val_loss: 10.4557 - val_acc: 0.3510\n",
      "Epoch 9/20\n",
      "6287/6287 [==============================] - 0s 19us/step - loss: 10.1357 - acc: 0.3366 - val_loss: 10.4484 - val_acc: 0.3518\n",
      "Epoch 10/20\n",
      "6287/6287 [==============================] - 0s 17us/step - loss: 10.0941 - acc: 0.3413 - val_loss: 10.4424 - val_acc: 0.3521\n",
      "Epoch 11/20\n",
      "6287/6287 [==============================] - 0s 18us/step - loss: 10.0804 - acc: 0.3405 - val_loss: 10.4424 - val_acc: 0.3521\n",
      "Epoch 12/20\n",
      "6287/6287 [==============================] - 0s 18us/step - loss: 9.9748 - acc: 0.3410 - val_loss: 10.4364 - val_acc: 0.3525\n",
      "Epoch 13/20\n",
      "6287/6287 [==============================] - 0s 18us/step - loss: 9.9103 - acc: 0.3405 - val_loss: 10.4304 - val_acc: 0.3529\n",
      "Epoch 14/20\n",
      "6287/6287 [==============================] - 0s 17us/step - loss: 9.7879 - acc: 0.3437 - val_loss: 10.4304 - val_acc: 0.3529\n",
      "Epoch 15/20\n",
      "6287/6287 [==============================] - 0s 18us/step - loss: 9.6743 - acc: 0.3420 - val_loss: 10.4304 - val_acc: 0.3529\n",
      "Epoch 16/20\n",
      "6287/6287 [==============================] - 0s 18us/step - loss: 9.6380 - acc: 0.3445 - val_loss: 10.4304 - val_acc: 0.3529\n",
      "Epoch 17/20\n",
      "6287/6287 [==============================] - 0s 18us/step - loss: 9.5407 - acc: 0.3423 - val_loss: 10.4304 - val_acc: 0.3529\n",
      "Epoch 18/20\n",
      "6287/6287 [==============================] - 0s 20us/step - loss: 9.5134 - acc: 0.3456 - val_loss: 10.4424 - val_acc: 0.3521\n",
      "Epoch 19/20\n",
      "6287/6287 [==============================] - 0s 23us/step - loss: 9.4455 - acc: 0.3452 - val_loss: 10.4424 - val_acc: 0.3521\n",
      "Epoch 20/20\n",
      "6287/6287 [==============================] - 0s 23us/step - loss: 9.4277 - acc: 0.3485 - val_loss: 10.4424 - val_acc: 0.3521\n"
     ]
    }
   ],
   "source": [
    "batch_size = 64 # number of examples per batch\n",
    "n_epochs = 20   # number of epochs to train - each epoch is a full pass in the entire trainset   \n",
    "validation_split = 0.3  # split from the trainset to be used to compute validation scores - the accuracy on an independent set of examples\n",
    "\n",
    "history = model.fit(input_train, y_train,\n",
    "                    epochs=n_epochs,\n",
    "                    batch_size=batch_size,\n",
    "                    validation_split=validation_split)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Performance Plot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "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"
    },
    {
     "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",
    "%matplotlib inline\n",
    "\n",
    "acc = history.history['acc']\n",
    "val_acc = history.history['val_acc']\n",
    "loss = history.history['loss']\n",
    "val_loss = history.history['val_loss']\n",
    "\n",
    "epochs = range(len(acc))\n",
    "\n",
    "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.legend()\n",
    "\n",
    "plt.figure()\n",
    "\n",
    "plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "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": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}