01_linear_regression.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Generate dataset\n",
    "trX = np.linspace(-1, 1, 101)\n",
    "trY = 2 * trX + np.random.randn(*trX.shape) * 0.33 # create a y value which is approximately linear but with some random noise"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Linear regression model\n",
    "model = Sequential()\n",
    "model.add(Dense(input_dim=1, output_dim=1, init='uniform', activation='linear'))\n",
    "model.compile(optimizer='sgd', loss='mse')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Linear regression model is initialized with weight w: -0.02, b: 0.00\n"
     ]
    }
   ],
   "source": [
    "# Print initial weights\n",
    "weights = model.layers[0].get_weights()\n",
    "w_init = weights[0][0][0]\n",
    "b_init = weights[1][0]\n",
    "print('Linear regression model is initialized with weight w: %.2f, b: %.2f' % (w_init, b_init))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "101/101 [==============================] - 0s - loss: 1.6034     \n",
      "Epoch 2/100\n",
      "101/101 [==============================] - 0s - loss: 1.5246     \n",
      "Epoch 3/100\n",
      "101/101 [==============================] - 0s - loss: 1.4446     \n",
      "Epoch 4/100\n",
      "101/101 [==============================] - 0s - loss: 1.3690     \n",
      "Epoch 5/100\n",
      "101/101 [==============================] - 0s - loss: 1.2975     \n",
      "Epoch 6/100\n",
      "101/101 [==============================] - 0s - loss: 1.2362     \n",
      "Epoch 7/100\n",
      "101/101 [==============================] - 0s - loss: 1.1691     \n",
      "Epoch 8/100\n",
      "101/101 [==============================] - 0s - loss: 1.1179     \n",
      "Epoch 9/100\n",
      "101/101 [==============================] - 0s - loss: 1.0717     \n",
      "Epoch 10/100\n",
      "101/101 [==============================] - 0s - loss: 1.0224     \n",
      "Epoch 11/100\n",
      "101/101 [==============================] - 0s - loss: 0.9743     \n",
      "Epoch 12/100\n",
      "101/101 [==============================] - 0s - loss: 0.9290     \n",
      "Epoch 13/100\n",
      "101/101 [==============================] - 0s - loss: 0.8886     \n",
      "Epoch 14/100\n",
      "101/101 [==============================] - 0s - loss: 0.8452     \n",
      "Epoch 15/100\n",
      "101/101 [==============================] - 0s - loss: 0.8061     \n",
      "Epoch 16/100\n",
      "101/101 [==============================] - 0s - loss: 0.7721     \n",
      "Epoch 17/100\n",
      "101/101 [==============================] - 0s - loss: 0.7335     \n",
      "Epoch 18/100\n",
      "101/101 [==============================] - 0s - loss: 0.6992     \n",
      "Epoch 19/100\n",
      "101/101 [==============================] - 0s - loss: 0.6663     \n",
      "Epoch 20/100\n",
      "101/101 [==============================] - 0s - loss: 0.6388     \n",
      "Epoch 21/100\n",
      "101/101 [==============================] - 0s - loss: 0.6106     \n",
      "Epoch 22/100\n",
      "101/101 [==============================] - 0s - loss: 0.5842     \n",
      "Epoch 23/100\n",
      "101/101 [==============================] - 0s - loss: 0.5584     \n",
      "Epoch 24/100\n",
      "101/101 [==============================] - 0s - loss: 0.5357     \n",
      "Epoch 25/100\n",
      "101/101 [==============================] - 0s - loss: 0.5126     \n",
      "Epoch 26/100\n",
      "101/101 [==============================] - 0s - loss: 0.4931     \n",
      "Epoch 27/100\n",
      "101/101 [==============================] - 0s - loss: 0.4715     \n",
      "Epoch 28/100\n",
      "101/101 [==============================] - 0s - loss: 0.4544     \n",
      "Epoch 29/100\n",
      "101/101 [==============================] - 0s - loss: 0.4364     \n",
      "Epoch 30/100\n",
      "101/101 [==============================] - 0s - loss: 0.4194     \n",
      "Epoch 31/100\n",
      "101/101 [==============================] - 0s - loss: 0.4032     \n",
      "Epoch 32/100\n",
      "101/101 [==============================] - 0s - loss: 0.3883     \n",
      "Epoch 33/100\n",
      "101/101 [==============================] - 0s - loss: 0.3703     \n",
      "Epoch 34/100\n",
      "101/101 [==============================] - 0s - loss: 0.3565     \n",
      "Epoch 35/100\n",
      "101/101 [==============================] - 0s - loss: 0.3433     \n",
      "Epoch 36/100\n",
      "101/101 [==============================] - 0s - loss: 0.3312     \n",
      "Epoch 37/100\n",
      "101/101 [==============================] - 0s - loss: 0.3179     \n",
      "Epoch 38/100\n",
      "101/101 [==============================] - 0s - loss: 0.3067     \n",
      "Epoch 39/100\n",
      "101/101 [==============================] - 0s - loss: 0.2956     \n",
      "Epoch 40/100\n",
      "101/101 [==============================] - 0s - loss: 0.2856     \n",
      "Epoch 41/100\n",
      "101/101 [==============================] - 0s - loss: 0.2767     \n",
      "Epoch 42/100\n",
      "101/101 [==============================] - 0s - loss: 0.2660     \n",
      "Epoch 43/100\n",
      "101/101 [==============================] - 0s - loss: 0.2579     \n",
      "Epoch 44/100\n",
      "101/101 [==============================] - 0s - loss: 0.2498     \n",
      "Epoch 45/100\n",
      "101/101 [==============================] - 0s - loss: 0.2428     \n",
      "Epoch 46/100\n",
      "101/101 [==============================] - 0s - loss: 0.2355     \n",
      "Epoch 47/100\n",
      "101/101 [==============================] - 0s - loss: 0.2281     \n",
      "Epoch 48/100\n",
      "101/101 [==============================] - 0s - loss: 0.2222     \n",
      "Epoch 49/100\n",
      "101/101 [==============================] - 0s - loss: 0.2158     \n",
      "Epoch 50/100\n",
      "101/101 [==============================] - 0s - loss: 0.2101     \n",
      "Epoch 51/100\n",
      "101/101 [==============================] - 0s - loss: 0.2054     \n",
      "Epoch 52/100\n",
      "101/101 [==============================] - 0s - loss: 0.2013     \n",
      "Epoch 53/100\n",
      "101/101 [==============================] - 0s - loss: 0.1963     \n",
      "Epoch 54/100\n",
      "101/101 [==============================] - 0s - loss: 0.1918     \n",
      "Epoch 55/100\n",
      "101/101 [==============================] - 0s - loss: 0.1870     \n",
      "Epoch 56/100\n",
      "101/101 [==============================] - 0s - loss: 0.1836     \n",
      "Epoch 57/100\n",
      "101/101 [==============================] - 0s - loss: 0.1800     \n",
      "Epoch 58/100\n",
      "101/101 [==============================] - 0s - loss: 0.1760     \n",
      "Epoch 59/100\n",
      "101/101 [==============================] - 0s - loss: 0.1721     \n",
      "Epoch 60/100\n",
      "101/101 [==============================] - 0s - loss: 0.1691     \n",
      "Epoch 61/100\n",
      "101/101 [==============================] - 0s - loss: 0.1664     \n",
      "Epoch 62/100\n",
      "101/101 [==============================] - 0s - loss: 0.1631     \n",
      "Epoch 63/100\n",
      "101/101 [==============================] - 0s - loss: 0.1602     \n",
      "Epoch 64/100\n",
      "101/101 [==============================] - 0s - loss: 0.1570     \n",
      "Epoch 65/100\n",
      "101/101 [==============================] - 0s - loss: 0.1542     \n",
      "Epoch 66/100\n",
      "101/101 [==============================] - 0s - loss: 0.1519     \n",
      "Epoch 67/100\n",
      "101/101 [==============================] - 0s - loss: 0.1496     \n",
      "Epoch 68/100\n",
      "101/101 [==============================] - 0s - loss: 0.1470     \n",
      "Epoch 69/100\n",
      "101/101 [==============================] - 0s - loss: 0.1446     \n",
      "Epoch 70/100\n",
      "101/101 [==============================] - 0s - loss: 0.1430     \n",
      "Epoch 71/100\n",
      "101/101 [==============================] - 0s - loss: 0.1414     \n",
      "Epoch 72/100\n",
      "101/101 [==============================] - 0s - loss: 0.1397     \n",
      "Epoch 73/100\n",
      "101/101 [==============================] - 0s - loss: 0.1381     \n",
      "Epoch 74/100\n",
      "101/101 [==============================] - 0s - loss: 0.1364     \n",
      "Epoch 75/100\n",
      "101/101 [==============================] - 0s - loss: 0.1352     \n",
      "Epoch 76/100\n",
      "101/101 [==============================] - 0s - loss: 0.1338     \n",
      "Epoch 77/100\n",
      "101/101 [==============================] - 0s - loss: 0.1321     \n",
      "Epoch 78/100\n",
      "101/101 [==============================] - 0s - loss: 0.1312     \n",
      "Epoch 79/100\n",
      "101/101 [==============================] - 0s - loss: 0.1299     \n",
      "Epoch 80/100\n",
      "101/101 [==============================] - 0s - loss: 0.1286     \n",
      "Epoch 81/100\n",
      "101/101 [==============================] - 0s - loss: 0.1270     \n",
      "Epoch 82/100\n",
      "101/101 [==============================] - 0s - loss: 0.1260     \n",
      "Epoch 83/100\n",
      "101/101 [==============================] - 0s - loss: 0.1252     \n",
      "Epoch 84/100\n",
      "101/101 [==============================] - 0s - loss: 0.1247     \n",
      "Epoch 85/100\n",
      "101/101 [==============================] - 0s - loss: 0.1239     \n",
      "Epoch 86/100\n",
      "101/101 [==============================] - 0s - loss: 0.1228     \n",
      "Epoch 87/100\n",
      "101/101 [==============================] - 0s - loss: 0.1222     \n",
      "Epoch 88/100\n",
      "101/101 [==============================] - 0s - loss: 0.1214     \n",
      "Epoch 89/100\n",
      "101/101 [==============================] - 0s - loss: 0.1210     \n",
      "Epoch 90/100\n",
      "101/101 [==============================] - 0s - loss: 0.1201     \n",
      "Epoch 91/100\n",
      "101/101 [==============================] - 0s - loss: 0.1194     \n",
      "Epoch 92/100\n",
      "101/101 [==============================] - 0s - loss: 0.1189     \n",
      "Epoch 93/100\n",
      "101/101 [==============================] - 0s - loss: 0.1184     \n",
      "Epoch 94/100\n",
      "101/101 [==============================] - 0s - loss: 0.1176     \n",
      "Epoch 95/100\n",
      "101/101 [==============================] - 0s - loss: 0.1168     \n",
      "Epoch 96/100\n",
      "101/101 [==============================] - 0s - loss: 0.1162     \n",
      "Epoch 97/100\n",
      "101/101 [==============================] - 0s - loss: 0.1157     \n",
      "Epoch 98/100\n",
      "101/101 [==============================] - 0s - loss: 0.1150     \n",
      "Epoch 99/100\n",
      "101/101 [==============================] - 0s - loss: 0.1144     \n",
      "Epoch 100/100\n",
      "101/101 [==============================] - 0s - loss: 0.1140     \n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7f9644059750>"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Train\n",
    "model.fit(trX, trY, nb_epoch=100, verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Linear regression model is trained with weight w: 1.95, b: -0.01\n"
     ]
    }
   ],
   "source": [
    "# Print trained weights\n",
    "weights = model.layers[0].get_weights()\n",
    "w = weights[0][0][0]\n",
    "b = weights[1][0]\n",
    "print('Linear regression model is trained with weight w: %.2f, b: %.2f' % (w, b))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAgoAAAFkCAYAAABB1xPiAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nOSiYmZlZLgcFMzMzy+WgYGZmZrn+H1ncfKqNadpdAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f9634402490>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(trX, trY, label='data')\n",
    "plt.plot(trX, w_init*trX + b_init, label='init')\n",
    "plt.plot(trX, w*trX + b, label='prediction')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3.0
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.4.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}