optimization.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Test of different gradient descent algorithms\n",
    "\n",
    "This jupyter notebook shows you the usage of different optimization techniques introduced in the class.\n",
    "\n",
    "\n",
    "We use these optimization algorithms to train the linear regression model over the house dataset.\n",
    "The following techniques will be tested\n",
    "1. Data normalization, which is effective to decrease the loss (i.e error).\n",
    "2. GD, SGD, mini-batch SGD, mini-batch SGD with momentum, RMSprop and Adam. The difference between GD and SGD is very significant, in terms of both loss value and efficiency. The difference for other algorithms is not very significant for this example because the model is very simple. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "# author: Yaqi Xie\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "from matplotlib import *\n",
    "from mpl_toolkits.mplot3d import Axes3D\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "from pylab import *\n",
    "import math\n",
    "import keras\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Activation\n",
    "from keras import optimizers\n",
    "%matplotlib inline\n",
    "\n",
    "# Read Data\n",
    "data_all = pd.read_csv('train.csv')\n",
    "data_x = data_all[['MSSubClass','LotArea','OverallQual','OverallCond','YearBuilt','YearRemodAdd','BsmtFinSF1','1stFlrSF','2ndFlrSF','GrLivArea','WoodDeckSF']]\n",
    "data_y = data_all['SalePrice']\n",
    "input_dim = data_x.shape[1]\n",
    "X = data_x.as_matrix()\n",
    "Y = data_y.as_matrix()\n",
    "total_len = len(X)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "1460/1460 [==============================] - 0s 34us/step - loss: 180549.9688\n",
      "Epoch 2/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 78974.1484\n",
      "Epoch 3/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 67284.5781\n",
      "Epoch 4/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 65136.8125\n",
      "Epoch 5/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 64288.7656\n",
      "Epoch 6/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 63617.5352\n",
      "Epoch 7/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 63010.8750\n",
      "Epoch 8/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 62413.0469\n",
      "Epoch 9/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 61819.7031\n",
      "Epoch 10/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 61235.4141\n",
      "Epoch 11/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 60656.9297\n",
      "Epoch 12/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 60090.9766\n",
      "Epoch 13/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 59537.1016\n",
      "Epoch 14/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 58988.5586\n",
      "Epoch 15/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 58447.6836\n",
      "Epoch 16/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 57908.4180\n",
      "Epoch 17/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 57372.4336\n",
      "Epoch 18/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 56845.2148\n",
      "Epoch 19/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 56325.5352\n",
      "Epoch 20/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 55814.2266\n",
      "Epoch 21/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 55317.6211\n",
      "Epoch 22/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 54823.2930\n",
      "Epoch 23/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 54343.4414\n",
      "Epoch 24/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 53877.8086\n",
      "Epoch 25/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 53420.3086\n",
      "Epoch 26/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 52967.2109\n",
      "Epoch 27/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 52524.1641\n",
      "Epoch 28/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 52094.1875\n",
      "Epoch 29/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 51674.4453\n",
      "Epoch 30/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 51274.1328\n",
      "Epoch 31/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 50895.4414\n",
      "Epoch 32/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 50532.1406\n",
      "Epoch 33/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 50174.3359\n",
      "Epoch 34/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 49838.3516\n",
      "Epoch 35/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 49498.9258\n",
      "Epoch 36/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 49147.6211\n",
      "Epoch 37/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 48852.4062\n",
      "Epoch 38/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 48550.0547\n",
      "Epoch 39/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 48262.3281\n",
      "Epoch 40/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 47983.8359\n",
      "Epoch 41/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 47705.0195\n",
      "Epoch 42/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 47435.8047\n",
      "Epoch 43/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 47174.5469\n",
      "Epoch 44/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 46920.0820\n",
      "Epoch 45/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 46676.9414\n",
      "Epoch 46/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 46447.5664\n",
      "Epoch 47/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 46226.6641\n",
      "Epoch 48/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 46007.2383\n",
      "Epoch 49/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 45790.3945\n",
      "Epoch 50/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 45580.1250\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/wangwei/miniconda3/lib/python3.6/site-packages/keras/models.py:939: UserWarning: The `nb_epoch` argument in `fit` has been renamed `epochs`.\n",
      "  warnings.warn('The `nb_epoch` argument in `fit` '\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7efcc40b9780>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# GD, 11 features, vector weights, without normalization\n",
    "model = Sequential([Dense(1, activation='linear', input_shape=(input_dim,))])\n",
    "gd = optimizers.SGD(lr=0.001, decay=0, momentum=0)  # please try 0.01\n",
    "model.compile(loss='mean_absolute_error', optimizer=gd)\n",
    "history = model.fit(X, Y, nb_epoch=50, batch_size=total_len)\n",
    "plt.plot(history.history['loss'])\n",
    "plt.xlabel('Iterations')\n",
    "plt.ylabel('Loss')\n",
    "plt.title('Gradient Descent \\n 11 feature (without normalization)')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "1460/1460 [==============================] - 0s 16us/step - loss: 0.3280\n",
      "Epoch 2/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3280\n",
      "Epoch 3/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3279\n",
      "Epoch 4/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3278\n",
      "Epoch 5/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3278\n",
      "Epoch 6/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3277\n",
      "Epoch 7/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3276\n",
      "Epoch 8/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3276\n",
      "Epoch 9/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3275\n",
      "Epoch 10/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3275\n",
      "Epoch 11/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3274\n",
      "Epoch 12/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3274\n",
      "Epoch 13/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3273\n",
      "Epoch 14/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3273\n",
      "Epoch 15/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3272\n",
      "Epoch 16/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3272\n",
      "Epoch 17/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3271\n",
      "Epoch 18/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3271\n",
      "Epoch 19/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3270\n",
      "Epoch 20/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3270\n",
      "Epoch 21/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3270\n",
      "Epoch 22/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3269\n",
      "Epoch 23/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3269\n",
      "Epoch 24/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3269\n",
      "Epoch 25/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3268\n",
      "Epoch 26/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3268\n",
      "Epoch 27/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3267\n",
      "Epoch 28/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3267\n",
      "Epoch 29/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3267\n",
      "Epoch 30/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3266\n",
      "Epoch 31/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3266\n",
      "Epoch 32/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3266\n",
      "Epoch 33/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3265\n",
      "Epoch 34/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3265\n",
      "Epoch 35/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3265\n",
      "Epoch 36/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3264\n",
      "Epoch 37/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3264\n",
      "Epoch 38/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3264\n",
      "Epoch 39/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3264\n",
      "Epoch 40/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3263\n",
      "Epoch 41/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3263\n",
      "Epoch 42/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3263\n",
      "Epoch 43/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3263\n",
      "Epoch 44/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3262\n",
      "Epoch 45/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3262\n",
      "Epoch 46/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3262\n",
      "Epoch 47/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3262\n",
      "Epoch 48/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3262\n",
      "Epoch 49/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3261\n",
      "Epoch 50/50\n",
      "1460/1460 [==============================] - 0s 1us/step - loss: 0.3261\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/wangwei/miniconda3/lib/python3.6/site-packages/keras/models.py:939: UserWarning: The `nb_epoch` argument in `fit` has been renamed `epochs`.\n",
      "  warnings.warn('The `nb_epoch` argument in `fit` '\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7efcc438fc50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#GD, 11 features, vector weights, with normalization \n",
    "data_x_normalized = (data_x - data_x.mean())*10 / (data_x.max() - data_x.min())\n",
    "data_y_normalized = (data_y - data_y.mean())*10 / (data_y.max() - data_y.min())\n",
    "X_normalized = data_x_normalized.as_matrix()\n",
    "Y_normalized = data_y_normalized.as_matrix()\n",
    "model2 = Sequential([Dense(1, activation='linear', input_shape=(input_dim,))])\n",
    "gd = optimizers.SGD(lr=0.01, decay=0, momentum=0)\n",
    "model2.compile(loss='mean_absolute_error', optimizer=gd)\n",
    "history = model.fit(X_normalized, Y_normalized, nb_epoch=50, batch_size=total_len)\n",
    "plt.plot(history.history['loss'])\n",
    "plt.xlabel('Iterations')\n",
    "plt.ylabel('Loss')\n",
    "plt.title('Gradient Descent \\n 11 feature (with normalization)')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Compare the above two figures for the effect of data normalization!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      " 321/1460 [=====>........................] - ETA: 0s - loss: 0.2919"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/wangwei/miniconda3/lib/python3.6/site-packages/keras/models.py:939: UserWarning: The `nb_epoch` argument in `fit` has been renamed `epochs`.\n",
      "  warnings.warn('The `nb_epoch` argument in `fit` '\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1460/1460 [==============================] - 1s 489us/step - loss: 0.3332\n",
      "Epoch 2/50\n",
      "1460/1460 [==============================] - 1s 483us/step - loss: 0.3332\n",
      "Epoch 3/50\n",
      "1460/1460 [==============================] - 1s 584us/step - loss: 0.3333\n",
      "Epoch 4/50\n",
      "1460/1460 [==============================] - 1s 655us/step - loss: 0.3332\n",
      "Epoch 5/50\n",
      "1460/1460 [==============================] - 1s 645us/step - loss: 0.3321\n",
      "Epoch 6/50\n",
      "1460/1460 [==============================] - 1s 622us/step - loss: 0.3335\n",
      "Epoch 7/50\n",
      "1460/1460 [==============================] - 1s 595us/step - loss: 0.3343\n",
      "Epoch 8/50\n",
      "1460/1460 [==============================] - 1s 536us/step - loss: 0.3312\n",
      "Epoch 9/50\n",
      "1460/1460 [==============================] - 1s 537us/step - loss: 0.3319\n",
      "Epoch 10/50\n",
      "1460/1460 [==============================] - 1s 548us/step - loss: 0.3306\n",
      "Epoch 11/50\n",
      "1460/1460 [==============================] - 1s 542us/step - loss: 0.3342\n",
      "Epoch 12/50\n",
      "1460/1460 [==============================] - 1s 536us/step - loss: 0.3306\n",
      "Epoch 13/50\n",
      "1460/1460 [==============================] - 1s 473us/step - loss: 0.3292\n",
      "Epoch 14/50\n",
      "1460/1460 [==============================] - 1s 482us/step - loss: 0.3327\n",
      "Epoch 15/50\n",
      "1460/1460 [==============================] - 1s 474us/step - loss: 0.3310\n",
      "Epoch 16/50\n",
      "1460/1460 [==============================] - 1s 472us/step - loss: 0.3307\n",
      "Epoch 17/50\n",
      "1460/1460 [==============================] - 1s 490us/step - loss: 0.3307\n",
      "Epoch 18/50\n",
      "1460/1460 [==============================] - 1s 515us/step - loss: 0.3319\n",
      "Epoch 19/50\n",
      "1460/1460 [==============================] - 1s 550us/step - loss: 0.3317\n",
      "Epoch 20/50\n",
      "1460/1460 [==============================] - 1s 540us/step - loss: 0.3319\n",
      "Epoch 21/50\n",
      "1460/1460 [==============================] - 1s 546us/step - loss: 0.3315\n",
      "Epoch 22/50\n",
      "1460/1460 [==============================] - 1s 537us/step - loss: 0.3302\n",
      "Epoch 23/50\n",
      "1460/1460 [==============================] - 1s 502us/step - loss: 0.3283\n",
      "Epoch 24/50\n",
      "1460/1460 [==============================] - 1s 474us/step - loss: 0.3292\n",
      "Epoch 25/50\n",
      "1460/1460 [==============================] - 1s 473us/step - loss: 0.3305\n",
      "Epoch 26/50\n",
      "1460/1460 [==============================] - 1s 490us/step - loss: 0.3305\n",
      "Epoch 27/50\n",
      "1460/1460 [==============================] - 1s 463us/step - loss: 0.3290\n",
      "Epoch 28/50\n",
      "1460/1460 [==============================] - 1s 477us/step - loss: 0.3290\n",
      "Epoch 29/50\n",
      "1460/1460 [==============================] - 1s 539us/step - loss: 0.3303\n",
      "Epoch 30/50\n",
      "1460/1460 [==============================] - 1s 545us/step - loss: 0.3284\n",
      "Epoch 31/50\n",
      "1460/1460 [==============================] - 1s 551us/step - loss: 0.3280\n",
      "Epoch 32/50\n",
      "1460/1460 [==============================] - 1s 551us/step - loss: 0.3302\n",
      "Epoch 33/50\n",
      "1460/1460 [==============================] - 1s 544us/step - loss: 0.3291\n",
      "Epoch 34/50\n",
      "1460/1460 [==============================] - 1s 489us/step - loss: 0.3274\n",
      "Epoch 35/50\n",
      "1460/1460 [==============================] - 1s 466us/step - loss: 0.3284\n",
      "Epoch 36/50\n",
      "1460/1460 [==============================] - 1s 480us/step - loss: 0.3307\n",
      "Epoch 37/50\n",
      "1460/1460 [==============================] - 1s 475us/step - loss: 0.3290\n",
      "Epoch 38/50\n",
      "1460/1460 [==============================] - 1s 479us/step - loss: 0.3274\n",
      "Epoch 39/50\n",
      "1460/1460 [==============================] - 1s 473us/step - loss: 0.3303\n",
      "Epoch 40/50\n",
      "1460/1460 [==============================] - 1s 576us/step - loss: 0.3270\n",
      "Epoch 41/50\n",
      "1460/1460 [==============================] - 1s 590us/step - loss: 0.3293\n",
      "Epoch 42/50\n",
      "1460/1460 [==============================] - 1s 550us/step - loss: 0.3296\n",
      "Epoch 43/50\n",
      "1460/1460 [==============================] - 1s 543us/step - loss: 0.3283\n",
      "Epoch 44/50\n",
      "1460/1460 [==============================] - 1s 545us/step - loss: 0.3279\n",
      "Epoch 45/50\n",
      "1460/1460 [==============================] - 1s 503us/step - loss: 0.3284\n",
      "Epoch 46/50\n",
      "1460/1460 [==============================] - 1s 501us/step - loss: 0.3290\n",
      "Epoch 47/50\n",
      "1460/1460 [==============================] - 1s 458us/step - loss: 0.3285\n",
      "Epoch 48/50\n",
      "1460/1460 [==============================] - 1s 467us/step - loss: 0.3276\n",
      "Epoch 49/50\n",
      "1460/1460 [==============================] - 1s 473us/step - loss: 0.3271\n",
      "Epoch 50/50\n",
      "1460/1460 [==============================] - 1s 509us/step - loss: 0.3270\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7efc9db75dd8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# SGD, 11 features, vector weights, with normalization\n",
    "data_x_normalized = (data_x - data_x.mean())*10 / (data_x.max() - data_x.min())\n",
    "data_y_normalized = (data_y - data_y.mean())*10 / (data_y.max() - data_y.min())\n",
    "X_normalized = data_x_normalized.as_matrix()\n",
    "Y_normalized = data_y_normalized.as_matrix()\n",
    "model2 = Sequential([Dense(1, activation='linear', input_shape=(input_dim,))])\n",
    "sgd = optimizers.SGD(lr=0.01, decay=0, momentum=0)\n",
    "model2.compile(loss='mean_absolute_error', optimizer=sgd)\n",
    "history = model.fit(X_normalized, Y_normalized, nb_epoch=50, batch_size=1)\n",
    "plt.plot(history.history['loss'])\n",
    "plt.xlabel('Iterations')\n",
    "plt.ylabel('Loss')\n",
    "plt.title('Stochastic Gradient Descent (Batch = 1) \\n 11 feature (with normalization)')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Compare the above two figures for GD and SGD"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "1460/1460 [==============================] - 0s 48us/step - loss: 0.3239\n",
      "Epoch 2/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3240\n",
      "Epoch 3/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3240\n",
      "Epoch 4/50\n",
      "  16/1460 [..............................] - ETA: 0s - loss: 0.1792"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/wangwei/miniconda3/lib/python3.6/site-packages/keras/models.py:939: UserWarning: The `nb_epoch` argument in `fit` has been renamed `epochs`.\n",
      "  warnings.warn('The `nb_epoch` argument in `fit` '\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3239\n",
      "Epoch 5/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3239\n",
      "Epoch 6/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3238\n",
      "Epoch 7/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3240\n",
      "Epoch 8/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3239\n",
      "Epoch 9/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3237\n",
      "Epoch 10/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3233\n",
      "Epoch 11/50\n",
      "1460/1460 [==============================] - 0s 33us/step - loss: 0.3239\n",
      "Epoch 12/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3238\n",
      "Epoch 13/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3239\n",
      "Epoch 14/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3241\n",
      "Epoch 15/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3237\n",
      "Epoch 16/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3241\n",
      "Epoch 17/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3236\n",
      "Epoch 18/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3238\n",
      "Epoch 19/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3238\n",
      "Epoch 20/50\n",
      "1460/1460 [==============================] - 0s 32us/step - loss: 0.3238\n",
      "Epoch 21/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3238\n",
      "Epoch 22/50\n",
      "1460/1460 [==============================] - 0s 33us/step - loss: 0.3239\n",
      "Epoch 23/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3240\n",
      "Epoch 24/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3238\n",
      "Epoch 25/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3239\n",
      "Epoch 26/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3238\n",
      "Epoch 27/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3238\n",
      "Epoch 28/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3238\n",
      "Epoch 29/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3241\n",
      "Epoch 30/50\n",
      "1460/1460 [==============================] - 0s 27us/step - loss: 0.3239\n",
      "Epoch 31/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3237\n",
      "Epoch 32/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3236\n",
      "Epoch 33/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3238\n",
      "Epoch 34/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3237\n",
      "Epoch 35/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3237\n",
      "Epoch 36/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3236\n",
      "Epoch 37/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3239\n",
      "Epoch 38/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3239\n",
      "Epoch 39/50\n",
      "1460/1460 [==============================] - 0s 34us/step - loss: 0.3236\n",
      "Epoch 40/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3236\n",
      "Epoch 41/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3236\n",
      "Epoch 42/50\n",
      "1460/1460 [==============================] - 0s 32us/step - loss: 0.3239\n",
      "Epoch 43/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3236\n",
      "Epoch 44/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3237\n",
      "Epoch 45/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3237\n",
      "Epoch 46/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3236\n",
      "Epoch 47/50\n",
      "1460/1460 [==============================] - 0s 33us/step - loss: 0.3234\n",
      "Epoch 48/50\n",
      "1460/1460 [==============================] - 0s 33us/step - loss: 0.3235\n",
      "Epoch 49/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3238\n",
      "Epoch 50/50\n",
      "1460/1460 [==============================] - 0s 32us/step - loss: 0.3235\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7efcc40c0400>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# minibatch SGD , 11 features, vector weights, with normalization\n",
    "data_x_normalized = (data_x - data_x.mean())*10 / (data_x.max() - data_x.min())\n",
    "data_y_normalized = (data_y - data_y.mean())*10 / (data_y.max() - data_y.min())\n",
    "X_normalized = data_x_normalized.as_matrix()\n",
    "Y_normalized = data_y_normalized.as_matrix()\n",
    "model2 = Sequential([Dense(1, activation='linear', input_shape=(input_dim,))])\n",
    "sgd2 = optimizers.SGD(lr=0.01, decay=0, momentum=0)\n",
    "model2.compile(loss='mean_absolute_error', optimizer=sgd)\n",
    "history = model.fit(X_normalized, Y_normalized, nb_epoch=50, batch_size=16)\n",
    "plt.plot(history.history['loss'])\n",
    "plt.xlabel('Iterations')\n",
    "plt.ylabel('Loss')\n",
    "plt.title('Mini-batch SGD  (Mini-Batch = 16) \\n 11 feature (with normalization)')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Please try different mini-batch size for the above mini-batch SGD!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "1460/1460 [==============================] - 0s 49us/step - loss: 0.3238\n",
      "Epoch 2/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3233\n",
      "Epoch 3/50\n",
      "1460/1460 [==============================] - 0s 33us/step - loss: 0.3236\n",
      "Epoch 4/50\n",
      "  16/1460 [..............................] - ETA: 0s - loss: 0.1926"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/wangwei/miniconda3/lib/python3.6/site-packages/keras/models.py:939: UserWarning: The `nb_epoch` argument in `fit` has been renamed `epochs`.\n",
      "  warnings.warn('The `nb_epoch` argument in `fit` '\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1460/1460 [==============================] - 0s 35us/step - loss: 0.3234\n",
      "Epoch 5/50\n",
      "1460/1460 [==============================] - 0s 33us/step - loss: 0.3234\n",
      "Epoch 6/50\n",
      "1460/1460 [==============================] - 0s 35us/step - loss: 0.3237\n",
      "Epoch 7/50\n",
      "1460/1460 [==============================] - 0s 33us/step - loss: 0.3235\n",
      "Epoch 8/50\n",
      "1460/1460 [==============================] - 0s 32us/step - loss: 0.3235\n",
      "Epoch 9/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3235\n",
      "Epoch 10/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3236\n",
      "Epoch 11/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3233\n",
      "Epoch 12/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3238\n",
      "Epoch 13/50\n",
      "1460/1460 [==============================] - 0s 33us/step - loss: 0.3235\n",
      "Epoch 14/50\n",
      "1460/1460 [==============================] - 0s 35us/step - loss: 0.3236\n",
      "Epoch 15/50\n",
      "1460/1460 [==============================] - 0s 33us/step - loss: 0.3237\n",
      "Epoch 16/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3235\n",
      "Epoch 17/50\n",
      "1460/1460 [==============================] - 0s 34us/step - loss: 0.3234\n",
      "Epoch 18/50\n",
      "1460/1460 [==============================] - 0s 34us/step - loss: 0.3231\n",
      "Epoch 19/50\n",
      "1460/1460 [==============================] - 0s 32us/step - loss: 0.3238\n",
      "Epoch 20/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3236\n",
      "Epoch 21/50\n",
      "1460/1460 [==============================] - 0s 33us/step - loss: 0.3233\n",
      "Epoch 22/50\n",
      "1460/1460 [==============================] - 0s 33us/step - loss: 0.3237\n",
      "Epoch 23/50\n",
      "1460/1460 [==============================] - 0s 32us/step - loss: 0.3234\n",
      "Epoch 24/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3235\n",
      "Epoch 25/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3238\n",
      "Epoch 26/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3233\n",
      "Epoch 27/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3235\n",
      "Epoch 28/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3233\n",
      "Epoch 29/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3234\n",
      "Epoch 30/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3232\n",
      "Epoch 31/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3235\n",
      "Epoch 32/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3235\n",
      "Epoch 33/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3234\n",
      "Epoch 34/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3235\n",
      "Epoch 35/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3234\n",
      "Epoch 36/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3235\n",
      "Epoch 37/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3232\n",
      "Epoch 38/50\n",
      "1460/1460 [==============================] - 0s 36us/step - loss: 0.3233\n",
      "Epoch 39/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3235\n",
      "Epoch 40/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3235\n",
      "Epoch 41/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3232\n",
      "Epoch 42/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3234\n",
      "Epoch 43/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3234\n",
      "Epoch 44/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3232\n",
      "Epoch 45/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3237\n",
      "Epoch 46/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3230\n",
      "Epoch 47/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3233\n",
      "Epoch 48/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3234\n",
      "Epoch 49/50\n",
      "1460/1460 [==============================] - 0s 34us/step - loss: 0.3229\n",
      "Epoch 50/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3235\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7efcc401d908>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# minibatch SGD - Momentum, 11 features, vector weights, with normalization\n",
    "data_x_normalized = (data_x - data_x.mean())*10 / (data_x.max() - data_x.min())\n",
    "data_y_normalized = (data_y - data_y.mean())*10 / (data_y.max() - data_y.min())\n",
    "X_normalized = data_x_normalized.as_matrix()\n",
    "Y_normalized = data_y_normalized.as_matrix()\n",
    "model2 = Sequential([Dense(1, activation='linear', input_shape=(input_dim,))])\n",
    "sgd2 = optimizers.SGD(lr=0.01, decay=0, momentum=0.9)\n",
    "model2.compile(loss='mean_absolute_error', optimizer=sgd)\n",
    "history = model.fit(X_normalized, Y_normalized, nb_epoch=50, batch_size=16)\n",
    "plt.plot(history.history['loss'])\n",
    "plt.xlabel('Iterations')\n",
    "plt.ylabel('Loss')\n",
    "plt.title('Mini-batch SGD with momentum (Mini-Batch = 16) \\n 11 feature (with normalization)')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "1460/1460 [==============================] - 0s 47us/step - loss: 0.3233\n",
      "Epoch 2/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3235\n",
      "Epoch 3/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3233\n",
      "Epoch 4/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3232\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/wangwei/miniconda3/lib/python3.6/site-packages/keras/models.py:939: UserWarning: The `nb_epoch` argument in `fit` has been renamed `epochs`.\n",
      "  warnings.warn('The `nb_epoch` argument in `fit` '\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 5/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3233\n",
      "Epoch 6/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3230\n",
      "Epoch 7/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3232\n",
      "Epoch 8/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3232\n",
      "Epoch 9/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3234\n",
      "Epoch 10/50\n",
      "1460/1460 [==============================] - 0s 27us/step - loss: 0.3232\n",
      "Epoch 11/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3231\n",
      "Epoch 12/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3233\n",
      "Epoch 13/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3231\n",
      "Epoch 14/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3234\n",
      "Epoch 15/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3232\n",
      "Epoch 16/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3233\n",
      "Epoch 17/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3230\n",
      "Epoch 18/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3231\n",
      "Epoch 19/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3230\n",
      "Epoch 20/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3233\n",
      "Epoch 21/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3231\n",
      "Epoch 22/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3234\n",
      "Epoch 23/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3234\n",
      "Epoch 24/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3230\n",
      "Epoch 25/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3232\n",
      "Epoch 26/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3232\n",
      "Epoch 27/50\n",
      "1460/1460 [==============================] - 0s 34us/step - loss: 0.3231\n",
      "Epoch 28/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3233\n",
      "Epoch 29/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3233\n",
      "Epoch 30/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3232\n",
      "Epoch 31/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3229\n",
      "Epoch 32/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3231\n",
      "Epoch 33/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3229\n",
      "Epoch 34/50\n",
      "1460/1460 [==============================] - 0s 32us/step - loss: 0.3234\n",
      "Epoch 35/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3230\n",
      "Epoch 36/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3231\n",
      "Epoch 37/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3231\n",
      "Epoch 38/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3230\n",
      "Epoch 39/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3230\n",
      "Epoch 40/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3233\n",
      "Epoch 41/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3231\n",
      "Epoch 42/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3232\n",
      "Epoch 43/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3230\n",
      "Epoch 44/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3231\n",
      "Epoch 45/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3230\n",
      "Epoch 46/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3230\n",
      "Epoch 47/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3230\n",
      "Epoch 48/50\n",
      "1460/1460 [==============================] - 0s 33us/step - loss: 0.3227\n",
      "Epoch 49/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3232\n",
      "Epoch 50/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3228\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7efc9dd7b748>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# minibatch SGD - RMSprop, 11 features, vector weights, with normalization \n",
    "data_x_normalized = (data_x - data_x.mean())*10 / (data_x.max() - data_x.min())\n",
    "data_y_normalized = (data_y - data_y.mean())*10 / (data_y.max() - data_y.min())\n",
    "X_normalized = data_x_normalized.as_matrix()\n",
    "Y_normalized = data_y_normalized.as_matrix()\n",
    "model2 = Sequential([Dense(1, activation='linear', input_shape=(input_dim,))])\n",
    "model2.compile(loss='mean_absolute_error', optimizer='RMSprop')\n",
    "history = model.fit(X_normalized, Y_normalized, nb_epoch=50, batch_size=16)\n",
    "plt.plot(history.history['loss'])\n",
    "plt.xlabel('Iterations')\n",
    "plt.ylabel('Loss')\n",
    "plt.title('RMSprop (Mini-Batch = 16) \\n 11 feature (with normalization) ')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "1460/1460 [==============================] - 0s 52us/step - loss: 0.3229\n",
      "Epoch 2/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3231\n",
      "Epoch 3/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3230\n",
      "Epoch 4/50\n",
      "  16/1460 [..............................] - ETA: 0s - loss: 0.1751"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/wangwei/miniconda3/lib/python3.6/site-packages/keras/models.py:939: UserWarning: The `nb_epoch` argument in `fit` has been renamed `epochs`.\n",
      "  warnings.warn('The `nb_epoch` argument in `fit` '\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3229\n",
      "Epoch 5/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3232\n",
      "Epoch 6/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3230\n",
      "Epoch 7/50\n",
      "1460/1460 [==============================] - 0s 34us/step - loss: 0.3229\n",
      "Epoch 8/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3230\n",
      "Epoch 9/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3230\n",
      "Epoch 10/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3228\n",
      "Epoch 11/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3230\n",
      "Epoch 12/50\n",
      "1460/1460 [==============================] - 0s 32us/step - loss: 0.3230\n",
      "Epoch 13/50\n",
      "1460/1460 [==============================] - 0s 33us/step - loss: 0.3229\n",
      "Epoch 14/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3229\n",
      "Epoch 15/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3228\n",
      "Epoch 16/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3225\n",
      "Epoch 17/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3232\n",
      "Epoch 18/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3230\n",
      "Epoch 19/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3229\n",
      "Epoch 20/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3230\n",
      "Epoch 21/50\n",
      "1460/1460 [==============================] - 0s 27us/step - loss: 0.3230\n",
      "Epoch 22/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3229\n",
      "Epoch 23/50\n",
      "1460/1460 [==============================] - 0s 27us/step - loss: 0.3230\n",
      "Epoch 24/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3225\n",
      "Epoch 25/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3229\n",
      "Epoch 26/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3229\n",
      "Epoch 27/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3229\n",
      "Epoch 28/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3229\n",
      "Epoch 29/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3229\n",
      "Epoch 30/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3230\n",
      "Epoch 31/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3229\n",
      "Epoch 32/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3227\n",
      "Epoch 33/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3227\n",
      "Epoch 34/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3228\n",
      "Epoch 35/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3225\n",
      "Epoch 36/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3227\n",
      "Epoch 37/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3226\n",
      "Epoch 38/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3228\n",
      "Epoch 39/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3226\n",
      "Epoch 40/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3226\n",
      "Epoch 41/50\n",
      "1460/1460 [==============================] - 0s 29us/step - loss: 0.3228\n",
      "Epoch 42/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3227\n",
      "Epoch 43/50\n",
      "1460/1460 [==============================] - 0s 28us/step - loss: 0.3226\n",
      "Epoch 44/50\n",
      "1460/1460 [==============================] - 0s 37us/step - loss: 0.3228\n",
      "Epoch 45/50\n",
      "1460/1460 [==============================] - 0s 32us/step - loss: 0.3228\n",
      "Epoch 46/50\n",
      "1460/1460 [==============================] - 0s 36us/step - loss: 0.3226\n",
      "Epoch 47/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3227\n",
      "Epoch 48/50\n",
      "1460/1460 [==============================] - 0s 31us/step - loss: 0.3225\n",
      "Epoch 49/50\n",
      "1460/1460 [==============================] - 0s 30us/step - loss: 0.3231\n",
      "Epoch 50/50\n",
      "1460/1460 [==============================] - 0s 32us/step - loss: 0.3226\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7efc9dcbeb70>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# minibatch SGD - Adam, 11 features, vector weights, with normalization \n",
    "data_x_normalized = (data_x - data_x.mean())*10 / (data_x.max() - data_x.min())\n",
    "data_y_normalized = (data_y - data_y.mean())*10 / (data_y.max() - data_y.min())\n",
    "X_normalized = data_x_normalized.as_matrix()\n",
    "Y_normalized = data_y_normalized.as_matrix()\n",
    "model2 = Sequential([Dense(1, activation='linear', input_shape=(input_dim,))])\n",
    "model2.compile(loss='mean_absolute_error', optimizer='Adam')\n",
    "history = model.fit(X_normalized, Y_normalized, nb_epoch=50, batch_size=16)\n",
    "plt.plot(history.history['loss'])\n",
    "plt.xlabel('Iterations')\n",
    "plt.ylabel('Loss')\n",
    "plt.title('Adam (Mini-Batch = 16) \\n 11 feature (with normalization)  ')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}

train.csv