Chapter05/Activity5.01/Activity5_01.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load The dataset\n",
    "import pandas as pd\n",
    "X = pd.read_csv('../data/avila-tr_feats.csv')\n",
    "y = pd.read_csv('../data/avila-tr_target.csv')\n",
    "\n",
    "# Split the dataset into training set and test set with a 0.8-0.2 ratio\n",
    "from sklearn.model_selection import train_test_split\n",
    "seed = 1\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=seed)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Accuracy on Validation Set = 0.8024927973747253\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# define a seed for random number generator so the result will be reproducible\n",
    "import numpy as np\n",
    "from tensorflow import random\n",
    "\n",
    "np.random.seed(seed)\n",
    "random.set_seed(seed)\n",
    "\n",
    "# define the keras model\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense\n",
    "model_1 = Sequential()\n",
    "model_1.add(Dense(10, input_dim=X_test.shape[1], activation='relu'))\n",
    "model_1.add(Dense(6, activation='relu'))\n",
    "model_1.add(Dense(4, activation='relu'))\n",
    "model_1.add(Dense(1, activation='sigmoid'))\n",
    "model_1.compile(loss='binary_crossentropy', optimizer='sgd', metrics=['accuracy'])\n",
    "# train the model using training set while evaluating on test set\n",
    "history=model_1.fit(X_train, y_train, batch_size = 20, epochs = 100, validation_data=(X_test, y_test), verbose=0, shuffle=False)\n",
    "\n",
    "# import require packages for plotting\n",
    "import matplotlib.pyplot as plt \n",
    "import matplotlib\n",
    "%matplotlib inline \n",
    "\n",
    "# plot training error and test error\n",
    "matplotlib.rcParams['figure.figsize'] = (10.0, 8.0) \n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.ylim(0,1)\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train loss', 'validation loss'], loc='upper right')\n",
    "\n",
    "# print the best accuracy reached on the test set\n",
    "print(\"Best Accuracy on Validation Set =\", max(history.history['val_accuracy']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Accuracy on Validation Set = 0.797698974609375\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# set up a seed for random number generator so the result will be reproducible\n",
    "np.random.seed(seed)\n",
    "random.set_seed(seed)\n",
    "\n",
    "# define the keras model with l2 regularization with lambda = 0.01\n",
    "from keras.regularizers import l2\n",
    "l2_param = 0.01\n",
    "model_2 = Sequential()\n",
    "model_2.add(Dense(10, input_dim=X_train.shape[1], activation='relu', kernel_regularizer=l2(l2_param)))\n",
    "model_2.add(Dense(6, activation='relu', kernel_regularizer=l2(l2_param)))\n",
    "model_2.add(Dense(4, activation='relu', kernel_regularizer=l2(l2_param)))\n",
    "model_2.add(Dense(1, activation='sigmoid'))\n",
    "model_2.compile(loss='binary_crossentropy', optimizer='sgd', metrics=['accuracy'])\n",
    "\n",
    "# train the model using training set while evaluationg on test set\n",
    "history=model_2.fit(X_train, y_train, batch_size = 20, epochs = 100, validation_data=(X_test, y_test), verbose=0, shuffle=False)\n",
    "\n",
    "# plot training error and test error\n",
    "matplotlib.rcParams['figure.figsize'] = (10.0, 8.0)\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.ylim(0,1)\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train loss', 'validation loss'], loc='upper right')\n",
    "# print the best accuracy reached on the test set\n",
    "print(\"Best Accuracy on Validation Set =\", max(history.history['val_accuracy']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Accuracy on Validation Set = 0.5910834074020386\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# set up a seed for random number generator so the result will be reproducible\n",
    "np.random.seed(seed)\n",
    "random.set_seed(seed)\n",
    "\n",
    "# define the keras model with l2 regularization with lambda = 0.1\n",
    "from keras.regularizers import l2\n",
    "l2_param = 0.1\n",
    "model_3 = Sequential()\n",
    "model_3.add(Dense(10, input_dim=X_train.shape[1], activation='relu', kernel_regularizer=l2(l2_param)))\n",
    "model_3.add(Dense(6, activation='relu', kernel_regularizer=l2(l2_param)))\n",
    "model_3.add(Dense(4, activation='relu', kernel_regularizer=l2(l2_param)))\n",
    "model_3.add(Dense(1, activation='sigmoid'))\n",
    "model_3.compile(loss='binary_crossentropy', optimizer='sgd', metrics=['accuracy'])\n",
    "\n",
    "# train the model using training set while evaluationg on test set\n",
    "history=model_3.fit(X_train, y_train, batch_size = 20, epochs = 100, validation_data=(X_test, y_test), verbose=0, shuffle=False)\n",
    "\n",
    "# plot training error and test error\n",
    "matplotlib.rcParams['figure.figsize'] = (10.0, 8.0)\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.ylim(0,1)\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train loss', 'validation loss'], loc='upper right')\n",
    "# print the best accuracy reached on the test set\n",
    "print(\"Best Accuracy on Validation Set =\", max(history.history['val_accuracy']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Accuracy on Validation Set = 0.8024927973747253\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# set up a seed for random number generator so the result will be reproducible\n",
    "np.random.seed(seed)\n",
    "random.set_seed(seed)\n",
    "\n",
    "# define the keras model with l2 regularization with lambda = 0.05\n",
    "from keras.regularizers import l2\n",
    "l2_param = 0.005\n",
    "model_4 = Sequential()\n",
    "model_4.add(Dense(10, input_dim=X_train.shape[1], activation='relu', kernel_regularizer=l2(l2_param)))\n",
    "model_4.add(Dense(6, activation='relu', kernel_regularizer=l2(l2_param)))\n",
    "model_4.add(Dense(4, activation='relu', kernel_regularizer=l2(l2_param)))\n",
    "model_4.add(Dense(1, activation='sigmoid'))\n",
    "model_4.compile(loss='binary_crossentropy', optimizer='sgd', metrics=['accuracy'])\n",
    "\n",
    "# train the model using training set while evaluationg on test set\n",
    "history=model_4.fit(X_train, y_train, batch_size = 20, epochs = 100, validation_data=(X_test, y_test), verbose=0, shuffle=False)\n",
    "\n",
    "# plot training error and test error\n",
    "matplotlib.rcParams['figure.figsize'] = (10.0, 8.0)\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.ylim(0,1)\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train loss', 'validation loss'], loc='upper right')\n",
    "# print the best accuracy reached on the test set\n",
    "print(\"Best Accuracy on Validation Set =\", max(history.history['val_accuracy']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Accuracy on Validation Set = 0.5910834074020386\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# set up a seed for random number generator so the result will be reproducible\n",
    "np.random.seed(seed)\n",
    "random.set_seed(seed)\n",
    "\n",
    "# define the keras model with l1 regularization with lambda = 0.01\n",
    "from keras.regularizers import l1\n",
    "l1_param = 0.01\n",
    "model_5 = Sequential()\n",
    "model_5.add(Dense(10, input_dim=X_train.shape[1], activation='relu', kernel_regularizer=l1(l1_param)))\n",
    "model_5.add(Dense(6, activation='relu', kernel_regularizer=l1(l1_param)))\n",
    "model_5.add(Dense(4, activation='relu', kernel_regularizer=l1(l1_param)))\n",
    "model_5.add(Dense(1, activation='sigmoid'))\n",
    "model_5.compile(loss='binary_crossentropy', optimizer='sgd', metrics=['accuracy'])\n",
    "\n",
    "# train the model using training set while evaluationg on test set\n",
    "history=model_5.fit(X_train, y_train, batch_size = 20, epochs = 100, validation_data=(X_test, y_test), verbose=0, shuffle=False)\n",
    "\n",
    "# plot training error and test error\n",
    "matplotlib.rcParams['figure.figsize'] = (10.0, 8.0)\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.ylim(0,1)\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train loss', 'validation loss'], loc='upper right')\n",
    "# print the best accuracy reached on the test set\n",
    "print(\"Best Accuracy on Validation Set =\", max(history.history['val_accuracy']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Accuracy on Validation Set = 0.7794822454452515\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# set up a seed for random number generator so the result will be reproducible\n",
    "np.random.seed(seed)\n",
    "random.set_seed(seed)\n",
    "\n",
    "# define the keras model with l1 regularization with lambda = 0.1\n",
    "from keras.regularizers import l1\n",
    "l1_param = 0.005\n",
    "model_6 = Sequential()\n",
    "model_6.add(Dense(10, input_dim=X_train.shape[1], activation='relu', kernel_regularizer=l1(l1_param)))\n",
    "model_6.add(Dense(6, activation='relu', kernel_regularizer=l1(l1_param)))\n",
    "model_6.add(Dense(4, activation='relu', kernel_regularizer=l1(l1_param)))\n",
    "model_6.add(Dense(1, activation='sigmoid'))\n",
    "model_6.compile(loss='binary_crossentropy', optimizer='sgd', metrics=['accuracy'])\n",
    "\n",
    "# train the model using training set while evaluationg on test set\n",
    "history=model_6.fit(X_train, y_train, batch_size = 20, epochs = 100, validation_data=(X_test, y_test), verbose=0, shuffle=False)\n",
    "\n",
    "# plot training error and test error\n",
    "matplotlib.rcParams['figure.figsize'] = (10.0, 8.0) \n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.ylim(0,1)\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train loss', 'validation loss'], loc='upper right')\n",
    "# print the best accuracy reached on the test set\n",
    "print(\"Best Accuracy on Validation Set =\", max(history.history['val_accuracy']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Accuracy on Validation Set = 0.5925215482711792\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# set up a seed for random number generator so the result will be reproducible\n",
    "np.random.seed(seed)\n",
    "random.set_seed(seed)\n",
    "\n",
    "# define the keras model with l1_l2 regularization with l1_lambda = 0.005 and l2_lambda = 0.005\n",
    "from keras.regularizers import l1_l2\n",
    "l1_param = 0.005\n",
    "l2_param = 0.005\n",
    "model_7 = Sequential()\n",
    "model_7.add(Dense(10, input_dim=X_train.shape[1], activation='relu', kernel_regularizer=l1_l2(l1=l1_param, l2=l2_param)))\n",
    "model_7.add(Dense(6, activation='relu', kernel_regularizer=l1_l2(l1=l1_param, l2=l2_param)))\n",
    "model_7.add(Dense(4, activation='relu', kernel_regularizer=l1_l2(l1=l1_param, l2=l2_param)))\n",
    "model_7.add(Dense(1, activation='sigmoid'))\n",
    "model_7.compile(loss='binary_crossentropy', optimizer='sgd', metrics=['accuracy'])\n",
    "\n",
    "# train the model using training set while evaluationg on test set\n",
    "history=model_7.fit(X_train, y_train, batch_size = 20, epochs = 100, validation_data=(X_test, y_test), verbose=0, shuffle=False)\n",
    " \n",
    "# plot training error and test error\n",
    "matplotlib.rcParams['figure.figsize'] = (10.0, 8.0)\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.ylim(0,1)\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train loss', 'validation loss'], loc='upper right')\n",
    "# print the best accuracy reached on the test set\n",
    "print(\"Best Accuracy on Validation Set =\", max(history.history['val_accuracy']))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "py3.7",
   "language": "python",
   "name": "py3.7"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}