keras_pg_85.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Boston housing pricing dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Importing dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.datasets import boston_housing\n",
    "import numpy as np\n",
    "\n",
    "(train_data, train_labels), (test_data, test_labels) = boston_housing.load_data()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Get a feel of the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training shape, number of examples --  (404, 13) , 404\n",
      "Test shape, number of examples --  (102, 13) , 102\n",
      "Example train data --  [  1.23247   0.        8.14      0.        0.538     6.142    91.7\n",
      "   3.9769    4.      307.       21.      396.9      18.72   ]\n",
      "Training label eg. --  15.2\n"
     ]
    }
   ],
   "source": [
    "print('Training shape, number of examples -- ', train_data.shape,',', len(train_data))\n",
    "print('Test shape, number of examples -- ', test_data.shape,',', len(test_data))\n",
    "print('Example train data -- ', train_data[0])\n",
    "print('Training label eg. -- ', train_labels[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Normalizing the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "mean = train_data.mean(axis=0)\n",
    "train_data -= mean\n",
    "std = train_data.std(axis=0)\n",
    "train_data /= std\n",
    "\n",
    "test_data -= mean\n",
    "test_data /= std"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Building the network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense\n",
    "\n",
    "def build_model(): # DRY objective\n",
    "    model = Sequential()\n",
    "    model.add(Dense(64, activation='relu', input_shape=(train_data.shape[1],)))\n",
    "    model.add(Dense(634, activation='relu'))\n",
    "    model.add(Dense(1))\n",
    "    model.compile(optimizer='adam', loss='mse', metrics=['mae'])\n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Using K-fold validation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<img src='https://i.stack.imgur.com/k76HN.png' height='40%' width='50%' alt='K-fold validation description'>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Processing fold # 0\n",
      "Specifications --  0 101\n",
      "Specificatinos --  (303, 13) (303,)\n",
      "Specifications -- #2 (101, 13) (101,)\n",
      "Train on 303 samples, validate on 101 samples\n",
      "Epoch 1/10\n",
      "303/303 [==============================] - 1s 3ms/step - loss: 200.8055 - mean_absolute_error: 10.4536 - val_loss: 122.4991 - val_mean_absolute_error: 7.8938\n",
      "Epoch 2/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 99.4145 - mean_absolute_error: 7.5584 - val_loss: 57.5995 - val_mean_absolute_error: 6.2220\n",
      "Epoch 3/10\n",
      "303/303 [==============================] - 0s 2ms/step - loss: 71.9582 - mean_absolute_error: 6.0305 - val_loss: 43.1951 - val_mean_absolute_error: 4.5437\n",
      "Epoch 4/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 62.1881 - mean_absolute_error: 5.7411 - val_loss: 46.4478 - val_mean_absolute_error: 5.4415\n",
      "Epoch 5/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 64.6251 - mean_absolute_error: 5.8561 - val_loss: 64.2330 - val_mean_absolute_error: 6.1223\n",
      "Epoch 6/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 57.7182 - mean_absolute_error: 5.4243 - val_loss: 51.1345 - val_mean_absolute_error: 5.7391\n",
      "Epoch 7/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 45.4541 - mean_absolute_error: 4.8515 - val_loss: 38.9662 - val_mean_absolute_error: 4.6079\n",
      "Epoch 8/10\n",
      "303/303 [==============================] - 0s 2ms/step - loss: 44.5485 - mean_absolute_error: 4.8814 - val_loss: 70.5292 - val_mean_absolute_error: 5.8003\n",
      "Epoch 9/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 51.8555 - mean_absolute_error: 5.2265 - val_loss: 33.8326 - val_mean_absolute_error: 4.0184\n",
      "Epoch 10/10\n",
      "303/303 [==============================] - 0s 2ms/step - loss: 41.8924 - mean_absolute_error: 4.7768 - val_loss: 38.6921 - val_mean_absolute_error: 4.4325\n",
      "Processing fold # 1\n",
      "Specifications --  101 202\n",
      "Specificatinos --  (303, 13) (303,)\n",
      "Specifications -- #2 (101, 13) (101,)\n",
      "Train on 303 samples, validate on 101 samples\n",
      "Epoch 1/10\n",
      "303/303 [==============================] - 1s 3ms/step - loss: 224.7131 - mean_absolute_error: 11.4034 - val_loss: 105.5477 - val_mean_absolute_error: 8.7163\n",
      "Epoch 2/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 102.7489 - mean_absolute_error: 7.3071 - val_loss: 66.9370 - val_mean_absolute_error: 6.1849\n",
      "Epoch 3/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 76.7475 - mean_absolute_error: 6.3653 - val_loss: 64.8548 - val_mean_absolute_error: 5.5864\n",
      "Epoch 4/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 68.2014 - mean_absolute_error: 6.0445 - val_loss: 61.0055 - val_mean_absolute_error: 5.4253\n",
      "Epoch 5/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 55.1394 - mean_absolute_error: 5.4517 - val_loss: 42.8512 - val_mean_absolute_error: 4.7102\n",
      "Epoch 6/10\n",
      "303/303 [==============================] - 1s 2ms/step - loss: 57.5301 - mean_absolute_error: 5.3650 - val_loss: 36.9560 - val_mean_absolute_error: 4.8805\n",
      "Epoch 7/10\n",
      "303/303 [==============================] - 1s 2ms/step - loss: 51.1423 - mean_absolute_error: 5.1556 - val_loss: 37.8625 - val_mean_absolute_error: 4.8098\n",
      "Epoch 8/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 43.4952 - mean_absolute_error: 4.7279 - val_loss: 29.5265 - val_mean_absolute_error: 3.8776\n",
      "Epoch 9/10\n",
      "303/303 [==============================] - 0s 2ms/step - loss: 46.5776 - mean_absolute_error: 4.8590 - val_loss: 39.9798 - val_mean_absolute_error: 4.5441\n",
      "Epoch 10/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 44.1465 - mean_absolute_error: 4.7335 - val_loss: 71.0943 - val_mean_absolute_error: 7.3088\n",
      "Processing fold # 2\n",
      "Specifications --  202 303\n",
      "Specificatinos --  (303, 13) (303,)\n",
      "Specifications -- #2 (101, 13) (101,)\n",
      "Train on 303 samples, validate on 101 samples\n",
      "Epoch 1/10\n",
      "303/303 [==============================] - 1s 3ms/step - loss: 211.7694 - mean_absolute_error: 10.5953 - val_loss: 77.0280 - val_mean_absolute_error: 6.3224\n",
      "Epoch 2/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 100.7768 - mean_absolute_error: 7.4086 - val_loss: 150.4066 - val_mean_absolute_error: 10.2793\n",
      "Epoch 3/10\n",
      "303/303 [==============================] - 0s 2ms/step - loss: 92.9669 - mean_absolute_error: 7.0726 - val_loss: 53.5627 - val_mean_absolute_error: 5.0422\n",
      "Epoch 4/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 79.9281 - mean_absolute_error: 6.6523 - val_loss: 61.3599 - val_mean_absolute_error: 6.5327\n",
      "Epoch 5/10\n",
      "303/303 [==============================] - 0s 2ms/step - loss: 59.5274 - mean_absolute_error: 5.5024 - val_loss: 51.6584 - val_mean_absolute_error: 5.4930\n",
      "Epoch 6/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 56.1776 - mean_absolute_error: 5.4752 - val_loss: 61.4717 - val_mean_absolute_error: 6.3742\n",
      "Epoch 7/10\n",
      "303/303 [==============================] - 0s 2ms/step - loss: 50.2342 - mean_absolute_error: 5.1550 - val_loss: 26.8816 - val_mean_absolute_error: 3.9080\n",
      "Epoch 8/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 43.2802 - mean_absolute_error: 4.8709 - val_loss: 23.4984 - val_mean_absolute_error: 3.4151\n",
      "Epoch 9/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 42.7920 - mean_absolute_error: 4.8850 - val_loss: 27.3124 - val_mean_absolute_error: 3.5651\n",
      "Epoch 10/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 40.0988 - mean_absolute_error: 4.6319 - val_loss: 40.1970 - val_mean_absolute_error: 4.6197\n",
      "Processing fold # 3\n",
      "Specifications --  303 404\n",
      "Specificatinos --  (303, 13) (303,)\n",
      "Specifications -- #2 (101, 13) (101,)\n",
      "Train on 303 samples, validate on 101 samples\n",
      "Epoch 1/10\n",
      "303/303 [==============================] - 1s 3ms/step - loss: 203.0314 - mean_absolute_error: 10.8115 - val_loss: 124.3150 - val_mean_absolute_error: 9.2090\n",
      "Epoch 2/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 152.5273 - mean_absolute_error: 9.6821 - val_loss: 116.3864 - val_mean_absolute_error: 7.1379\n",
      "Epoch 3/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 89.0387 - mean_absolute_error: 6.9798 - val_loss: 102.3268 - val_mean_absolute_error: 8.6834\n",
      "Epoch 4/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 101.8162 - mean_absolute_error: 7.2475 - val_loss: 88.1304 - val_mean_absolute_error: 6.0708\n",
      "Epoch 5/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 63.2488 - mean_absolute_error: 6.0470 - val_loss: 81.5587 - val_mean_absolute_error: 6.3703\n",
      "Epoch 6/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 51.1959 - mean_absolute_error: 5.1066 - val_loss: 71.5856 - val_mean_absolute_error: 5.3633\n",
      "Epoch 7/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 45.9350 - mean_absolute_error: 4.9749 - val_loss: 68.9156 - val_mean_absolute_error: 5.3756\n",
      "Epoch 8/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 49.1970 - mean_absolute_error: 5.2398 - val_loss: 78.2755 - val_mean_absolute_error: 5.6967\n",
      "Epoch 9/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 41.8115 - mean_absolute_error: 4.8561 - val_loss: 78.7632 - val_mean_absolute_error: 7.7763\n",
      "Epoch 10/10\n",
      "303/303 [==============================] - 0s 1ms/step - loss: 42.7324 - mean_absolute_error: 4.8000 - val_loss: 68.4668 - val_mean_absolute_error: 5.1914\n"
     ]
    }
   ],
   "source": [
    "k = 4\n",
    "num_epochs = 10\n",
    "all_mae_histories = []\n",
    "num_samples = len(train_data) // k\n",
    "\n",
    "for i in range(k):\n",
    "    print('Processing fold #', i)\n",
    "    val_data = train_data[i*num_samples: (i+1)*num_samples] # Slice out the subsection\n",
    "    val_labels = train_labels[i*num_samples: (i+1)*num_samples]\n",
    "    partial_x_train = np.concatenate([train_data[:i*num_samples], train_data[(i+1)*num_samples:]], axis=0)\n",
    "    partial_y_train = np.concatenate([train_labels[:i*num_samples], train_labels[(i+1)*num_samples:]], axis=0)\n",
    "    \n",
    "    model = build_model()\n",
    "    print('Specifications -- ', i*num_samples, (i+1)*num_samples)\n",
    "    print('Specificatinos -- ', partial_x_train.shape, partial_y_train.shape)\n",
    "    print('Specifications -- #2', val_data.shape, val_labels.shape)\n",
    "\n",
    "\n",
    "    history = model.fit(partial_x_train, partial_y_train, validation_data=(val_data, val_labels), epochs=num_epochs, batch_size=1, verbose=1)\n",
    "    mae_history = history.history['val_mean_absolute_error']\n",
    "    all_mae_histories.append(mae_history)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Building history of succesive mean K-fold validation scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "average_mae_history = [\n",
    "    np.mean([x[i] for x in all_mae_histories]) for i in range(num_epochs)\n",
    "]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Plotting validation scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "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",
    "plt.plot(range(1, len(average_mae_history) + 1), average_mae_history)\n",
    "plt.xlabel('Number of epochs')\n",
    "plt.ylabel('Validation MAE')\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
}