3.6_Reuters_example.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Classify Reuters newswires into 46 mutually exclusive topics\n",
    "This is a single label, multiclass classification.\n",
    "\n",
    "Deep Learning with Python by Francois Chollet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.datasets import reuters\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/reuters.npz\n",
      "2113536/2110848 [==============================] - 0s 0us/step\n"
     ]
    }
   ],
   "source": [
    "# The argument 'num_words=10000' restricts the data to the 10,000 msot frequently occurring words found in the data.\n",
    "(train_data, train_labels), (test_data, test_labels) = reuters.load_data(num_words=10000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "8982\n",
      "2246\n"
     ]
    }
   ],
   "source": [
    "# 8,982 training examples and 2,246 test examples\n",
    "print(len(train_data))\n",
    "print(len(test_data))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preparing the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def vectorize_sequences(sequences, dimension=10000):\n",
    "    re = np.zeros((len(sequences), dimension))\n",
    "    for i, sequence in enumerate(sequences):\n",
    "        re[i, sequence] = 1.\n",
    "    return re.astype('float32')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train = vectorize_sequences(train_data)\n",
    "x_test = vectorize_sequences(test_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# There are multiple possible labels, and each news article is labeled with only one.\n",
    "# Use one-hot encoding.\n",
    "from keras.utils.np_utils import to_categorical\n",
    "one_hot_train_labels = to_categorical(train_labels)\n",
    "one_hot_test_labels = to_categorical(test_labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Building the network\n",
    "There are $46$ possible categories. So we want to use layers with more than $46$ units.\\\n",
    "Let's go with $64$ units per layer."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras import models\n",
    "from keras import layers\n",
    "\n",
    "'''\n",
    "layers with: 10000[=input], 64, 64, 46[=output]\n",
    "For each sample, the network will output a 46-dimensional vector.\n",
    "'''\n",
    "model = models.Sequential()\n",
    "model.add(layers.Dense(64, activation='relu', input_shape=(10000,)))\n",
    "model.add(layers.Dense(64, activation='relu'))\n",
    "model.add(layers.Dense(46, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(optimizer='rmsprop', \n",
    "              loss='categorical_crossentropy', \n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training & Validation\n",
    "Set apart 1,000 samples in the training data to use as a validation set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_validation = x_train[:1000]\n",
    "x_training = x_train[1000:]\n",
    "\n",
    "y_validation = one_hot_train_labels[:1000]\n",
    "y_training = one_hot_train_labels[1000:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "16/16 [==============================] - 0s 19ms/step - loss: 0.1079 - accuracy: 0.9578 - val_loss: 1.1541 - val_accuracy: 0.7940\n",
      "Epoch 2/20\n",
      "16/16 [==============================] - 0s 17ms/step - loss: 0.1069 - accuracy: 0.9599 - val_loss: 1.1123 - val_accuracy: 0.7960\n",
      "Epoch 3/20\n",
      "16/16 [==============================] - 0s 17ms/step - loss: 0.1040 - accuracy: 0.9593 - val_loss: 1.1284 - val_accuracy: 0.7910\n",
      "Epoch 4/20\n",
      "16/16 [==============================] - 0s 16ms/step - loss: 0.1004 - accuracy: 0.9585 - val_loss: 1.1922 - val_accuracy: 0.7940\n",
      "Epoch 5/20\n",
      "16/16 [==============================] - 0s 17ms/step - loss: 0.1051 - accuracy: 0.9587 - val_loss: 1.1686 - val_accuracy: 0.7930\n",
      "Epoch 6/20\n",
      "16/16 [==============================] - 0s 17ms/step - loss: 0.1001 - accuracy: 0.9570 - val_loss: 1.1957 - val_accuracy: 0.7920\n",
      "Epoch 7/20\n",
      "16/16 [==============================] - 0s 18ms/step - loss: 0.1010 - accuracy: 0.9584 - val_loss: 1.1504 - val_accuracy: 0.8030\n",
      "Epoch 8/20\n",
      "16/16 [==============================] - 0s 17ms/step - loss: 0.0994 - accuracy: 0.9559 - val_loss: 1.2166 - val_accuracy: 0.7960\n",
      "Epoch 9/20\n",
      "16/16 [==============================] - 0s 18ms/step - loss: 0.0954 - accuracy: 0.9580 - val_loss: 1.1976 - val_accuracy: 0.7910\n",
      "Epoch 10/20\n",
      "16/16 [==============================] - 0s 17ms/step - loss: 0.0958 - accuracy: 0.9582 - val_loss: 1.2134 - val_accuracy: 0.7980\n",
      "Epoch 11/20\n",
      "16/16 [==============================] - 0s 17ms/step - loss: 0.0923 - accuracy: 0.9602 - val_loss: 1.2216 - val_accuracy: 0.7960\n",
      "Epoch 12/20\n",
      "16/16 [==============================] - 0s 17ms/step - loss: 0.0948 - accuracy: 0.9575 - val_loss: 1.2115 - val_accuracy: 0.7980\n",
      "Epoch 13/20\n",
      "16/16 [==============================] - 0s 17ms/step - loss: 0.0953 - accuracy: 0.9560 - val_loss: 1.2298 - val_accuracy: 0.7940\n",
      "Epoch 14/20\n",
      "16/16 [==============================] - 0s 16ms/step - loss: 0.0900 - accuracy: 0.9605 - val_loss: 1.3061 - val_accuracy: 0.7780\n",
      "Epoch 15/20\n",
      "16/16 [==============================] - 0s 17ms/step - loss: 0.0901 - accuracy: 0.9587 - val_loss: 1.2616 - val_accuracy: 0.7860\n",
      "Epoch 16/20\n",
      "16/16 [==============================] - 0s 18ms/step - loss: 0.0922 - accuracy: 0.9590 - val_loss: 1.2605 - val_accuracy: 0.7900\n",
      "Epoch 17/20\n",
      "16/16 [==============================] - 0s 17ms/step - loss: 0.0897 - accuracy: 0.9578 - val_loss: 1.3210 - val_accuracy: 0.7870\n",
      "Epoch 18/20\n",
      "16/16 [==============================] - 0s 17ms/step - loss: 0.0895 - accuracy: 0.9594 - val_loss: 1.2901 - val_accuracy: 0.7850\n",
      "Epoch 19/20\n",
      "16/16 [==============================] - 0s 17ms/step - loss: 0.0868 - accuracy: 0.9618 - val_loss: 1.3135 - val_accuracy: 0.7870\n",
      "Epoch 20/20\n",
      "16/16 [==============================] - 0s 15ms/step - loss: 0.0917 - accuracy: 0.9569 - val_loss: 1.2539 - val_accuracy: 0.7900\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(x_training, y_training, \n",
    "                     epochs=20, batch_size=512,\n",
    "                     validation_data=(x_validation, y_validation))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "loss_values = history.history['loss']               # training loss\n",
    "val_loss_values = history.history['val_loss']       # validation loss\n",
    "acc_values = history.history['accuracy']            # training accuracy \n",
    "val_acc_values = history.history['val_accuracy']    # validation accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "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",
    "epochs = range(1, len(acc_values) +1)\n",
    "plt.plot(epochs, loss_values, 'bo', label='Training loss')\n",
    "plt.plot(epochs, val_loss_values, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "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",
    "plt.plot(epochs, acc_values, 'bo', label='Training accuracy')\n",
    "plt.plot(epochs, val_acc_values, 'b', label='Validation accuracy')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('accuracy')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "71/71 [==============================] - 0s 2ms/step - loss: 1.4504 - accuracy: 0.7858\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[1.450442910194397, 0.7858415246009827]"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# evaludate() returns [loss, accuracy]\n",
    "model.evaluate(x_test, one_hot_test_labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Use a trained network to predict\n",
    "Use a trained network to generate predictions on new data\\\n",
    "For multiclass classification, the predicted category is the one with largest probability."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "prediction = model.predict(x_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1.0"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# sum of probability distribution is one.\n",
    "np.sum(prediction[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.98978615\n",
      "3\n"
     ]
    }
   ],
   "source": [
    "print(np.max(prediction[0]))\n",
    "print(np.argmax(prediction[0]))"
   ]
  },
  {
   "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": 4
}