Activation Functions.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "import keras\n",
    "from keras.datasets import mnist\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout\n",
    "from keras.optimizers import RMSprop\n",
    "from keras.layers.advanced_activations import LeakyReLU"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# Load the data\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train size: 60000\n",
      "Test size: 10000\n"
     ]
    }
   ],
   "source": [
    "(x_train, y_train), (x_test, y_test) = mnist.load_data()\n",
    "\n",
    "num_classes = 10\n",
    "x_train = x_train.reshape(60000, 784)\n",
    "x_test = x_test.reshape(10000, 784)\n",
    "x_train = x_train.astype('float32')\n",
    "x_test = x_test.astype('float32')\n",
    "x_train /= 255\n",
    "x_test /= 255\n",
    "y_train = keras.utils.to_categorical(y_train, num_classes)\n",
    "y_test = keras.utils.to_categorical(y_test, num_classes)\n",
    "\n",
    "\n",
    "print 'Train size:', x_train.shape[0]\n",
    "print 'Test size:', x_test.shape[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# Build the network "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dense_1 (None, 784) ==> (None, 512)\n",
      "dense_2 (None, 512) ==> (None, 10)\n",
      "\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_1 (Dense)              (None, 512)               401920    \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                5130      \n",
      "=================================================================\n",
      "Total params: 407,050\n",
      "Trainable params: 407,050\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Dense(512, activation='relu', input_shape=(784,)))\n",
    "model.add(Dense(num_classes, activation='softmax'))\n",
    "\n",
    "for l in model.layers:\n",
    "    print l.name, l.input_shape,'==>',l.output_shape\n",
    "print\n",
    "print model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Train the network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/5\n",
      "60000/60000 [==============================] - 3s - loss: 0.2606 - acc: 0.9243 - val_loss: 0.1304 - val_acc: 0.9604\n",
      "Epoch 2/5\n",
      "60000/60000 [==============================] - 3s - loss: 0.1039 - acc: 0.9693 - val_loss: 0.0967 - val_acc: 0.9695\n",
      "Epoch 3/5\n",
      "60000/60000 [==============================] - 3s - loss: 0.0683 - acc: 0.9792 - val_loss: 0.0752 - val_acc: 0.9758\n",
      "Epoch 4/5\n",
      "60000/60000 [==============================] - 3s - loss: 0.0497 - acc: 0.9856 - val_loss: 0.0723 - val_acc: 0.9782\n",
      "Epoch 5/5\n",
      "60000/60000 [==============================] - 3s - loss: 0.0373 - acc: 0.9885 - val_loss: 0.0625 - val_acc: 0.9809\n",
      "\n",
      "Test loss: 0.062\n",
      "Test accuracy: 0.981\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x11ba84a90>"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12d69ef50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "batch_size = 128\n",
    "epochs = 5\n",
    "\n",
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer=RMSprop(),\n",
    "              metrics=['accuracy'])\n",
    "    \n",
    "history = model.fit(x_train, y_train,\n",
    "                    batch_size=batch_size,\n",
    "                    epochs=epochs,\n",
    "                    verbose=1,\n",
    "                    validation_data=(x_test, y_test))\n",
    "\n",
    "score = model.evaluate(x_test, y_test, verbose=100)\n",
    "\n",
    "print\n",
    "print 'Test loss:', round(score[0], 3)\n",
    "print 'Test accuracy:', round(score[1], 3)\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Build networks using all activations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.3870 - acc: 0.8884 - val_loss: 0.3389 - val_acc: 0.9049\n",
      "Epoch 2/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.3099 - acc: 0.9128 - val_loss: 0.2919 - val_acc: 0.9164\n",
      "Epoch 3/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2969 - acc: 0.9176 - val_loss: 0.3060 - val_acc: 0.9178\n",
      "Epoch 4/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2888 - acc: 0.9192 - val_loss: 0.3191 - val_acc: 0.9096\n",
      "Epoch 5/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2824 - acc: 0.9221 - val_loss: 0.3028 - val_acc: 0.9171\n",
      "Epoch 6/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2782 - acc: 0.9228 - val_loss: 0.2911 - val_acc: 0.9227\n",
      "Epoch 7/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2756 - acc: 0.9234 - val_loss: 0.2802 - val_acc: 0.9229\n",
      "Epoch 8/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2726 - acc: 0.9238 - val_loss: 0.2837 - val_acc: 0.9240\n",
      "Epoch 9/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2701 - acc: 0.9241 - val_loss: 0.2899 - val_acc: 0.9231\n",
      "Epoch 10/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2687 - acc: 0.9252 - val_loss: 0.2841 - val_acc: 0.9234\n",
      "Epoch 11/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2664 - acc: 0.9257 - val_loss: 0.2834 - val_acc: 0.9242\n",
      "Epoch 12/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2666 - acc: 0.9265 - val_loss: 0.2867 - val_acc: 0.9212\n",
      "Epoch 13/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2641 - acc: 0.9271 - val_loss: 0.3022 - val_acc: 0.9160\n",
      "Epoch 14/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2618 - acc: 0.9273 - val_loss: 0.3103 - val_acc: 0.9177\n",
      "Epoch 15/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2615 - acc: 0.9274 - val_loss: 0.2843 - val_acc: 0.9261\n",
      "Epoch 16/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2606 - acc: 0.9276 - val_loss: 0.2914 - val_acc: 0.9228\n",
      "Epoch 17/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2596 - acc: 0.9285 - val_loss: 0.2813 - val_acc: 0.9251\n",
      "Epoch 18/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2583 - acc: 0.9290 - val_loss: 0.2990 - val_acc: 0.9174\n",
      "Epoch 19/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2586 - acc: 0.9286 - val_loss: 0.2943 - val_acc: 0.9187\n",
      "Epoch 20/20\n",
      "60000/60000 [==============================] - 2s - loss: 0.2571 - acc: 0.9286 - val_loss: 0.2854 - val_acc: 0.9245\n",
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.4253 - acc: 0.8849 - val_loss: 0.2585 - val_acc: 0.9245\n",
      "Epoch 2/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2342 - acc: 0.9314 - val_loss: 0.2117 - val_acc: 0.9389\n",
      "Epoch 3/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.1774 - acc: 0.9474 - val_loss: 0.1583 - val_acc: 0.9535\n",
      "Epoch 4/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.1392 - acc: 0.9595 - val_loss: 0.1268 - val_acc: 0.9630\n",
      "Epoch 5/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.1123 - acc: 0.9679 - val_loss: 0.1129 - val_acc: 0.9666\n",
      "Epoch 6/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0933 - acc: 0.9729 - val_loss: 0.1032 - val_acc: 0.9697\n",
      "Epoch 7/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0788 - acc: 0.9769 - val_loss: 0.0938 - val_acc: 0.9707\n",
      "Epoch 8/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0676 - acc: 0.9800 - val_loss: 0.0893 - val_acc: 0.9734\n",
      "Epoch 9/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0579 - acc: 0.9832 - val_loss: 0.0779 - val_acc: 0.9763\n",
      "Epoch 10/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0507 - acc: 0.9853 - val_loss: 0.0751 - val_acc: 0.9776\n",
      "Epoch 11/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0442 - acc: 0.9874 - val_loss: 0.0697 - val_acc: 0.9793\n",
      "Epoch 12/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0386 - acc: 0.9891 - val_loss: 0.0716 - val_acc: 0.9788\n",
      "Epoch 13/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0336 - acc: 0.9907 - val_loss: 0.0689 - val_acc: 0.9797\n",
      "Epoch 14/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0295 - acc: 0.9918 - val_loss: 0.0676 - val_acc: 0.9796\n",
      "Epoch 15/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0256 - acc: 0.9932 - val_loss: 0.0671 - val_acc: 0.9793\n",
      "Epoch 16/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0225 - acc: 0.9941 - val_loss: 0.0633 - val_acc: 0.9815\n",
      "Epoch 17/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0194 - acc: 0.9953 - val_loss: 0.0655 - val_acc: 0.9811\n",
      "Epoch 18/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0172 - acc: 0.9957 - val_loss: 0.0634 - val_acc: 0.9809\n",
      "Epoch 19/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0149 - acc: 0.9964 - val_loss: 0.0652 - val_acc: 0.9818\n",
      "Epoch 20/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0129 - acc: 0.9974 - val_loss: 0.0626 - val_acc: 0.9826\n",
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.3327 - acc: 0.9018 - val_loss: 0.2167 - val_acc: 0.9359\n",
      "Epoch 2/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.1752 - acc: 0.9492 - val_loss: 0.1289 - val_acc: 0.9637\n",
      "Epoch 3/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.1159 - acc: 0.9663 - val_loss: 0.1141 - val_acc: 0.9671\n",
      "Epoch 4/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0849 - acc: 0.9753 - val_loss: 0.0899 - val_acc: 0.9732\n",
      "Epoch 5/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0649 - acc: 0.9803 - val_loss: 0.0844 - val_acc: 0.9720\n",
      "Epoch 6/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0508 - acc: 0.9848 - val_loss: 0.0712 - val_acc: 0.9779\n",
      "Epoch 7/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0400 - acc: 0.9884 - val_loss: 0.0721 - val_acc: 0.9782\n",
      "Epoch 8/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0323 - acc: 0.9908 - val_loss: 0.0658 - val_acc: 0.9806\n",
      "Epoch 9/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0259 - acc: 0.9925 - val_loss: 0.0655 - val_acc: 0.9808\n",
      "Epoch 10/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0204 - acc: 0.9943 - val_loss: 0.0662 - val_acc: 0.9794\n",
      "Epoch 11/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0165 - acc: 0.9958 - val_loss: 0.0686 - val_acc: 0.9810\n",
      "Epoch 12/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0128 - acc: 0.9967 - val_loss: 0.0617 - val_acc: 0.9820\n",
      "Epoch 13/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0103 - acc: 0.9975 - val_loss: 0.0697 - val_acc: 0.9790\n",
      "Epoch 14/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0078 - acc: 0.9982 - val_loss: 0.0674 - val_acc: 0.9824\n",
      "Epoch 15/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0063 - acc: 0.9986 - val_loss: 0.0665 - val_acc: 0.9817\n",
      "Epoch 16/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0049 - acc: 0.9990 - val_loss: 0.0656 - val_acc: 0.9821\n",
      "Epoch 17/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0041 - acc: 0.9992 - val_loss: 0.0701 - val_acc: 0.9821\n",
      "Epoch 18/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0031 - acc: 0.9993 - val_loss: 0.0700 - val_acc: 0.9825\n",
      "Epoch 19/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0026 - acc: 0.9996 - val_loss: 0.0701 - val_acc: 0.9825\n",
      "Epoch 20/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0022 - acc: 0.9996 - val_loss: 0.0761 - val_acc: 0.9819\n",
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.2598 - acc: 0.9252 - val_loss: 0.1312 - val_acc: 0.9591\n",
      "Epoch 2/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.1043 - acc: 0.9694 - val_loss: 0.0941 - val_acc: 0.9704\n",
      "Epoch 3/20\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "60000/60000 [==============================] - 3s - loss: 0.0687 - acc: 0.9792 - val_loss: 0.0891 - val_acc: 0.9737\n",
      "Epoch 4/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0493 - acc: 0.9848 - val_loss: 0.0680 - val_acc: 0.9792\n",
      "Epoch 5/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0372 - acc: 0.9886 - val_loss: 0.0744 - val_acc: 0.9789\n",
      "Epoch 6/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0282 - acc: 0.9916 - val_loss: 0.0808 - val_acc: 0.9764\n",
      "Epoch 7/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0221 - acc: 0.9931 - val_loss: 0.0659 - val_acc: 0.9821\n",
      "Epoch 8/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0174 - acc: 0.9948 - val_loss: 0.0690 - val_acc: 0.9819\n",
      "Epoch 9/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0130 - acc: 0.9961 - val_loss: 0.0709 - val_acc: 0.9825\n",
      "Epoch 10/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0103 - acc: 0.9972 - val_loss: 0.0712 - val_acc: 0.9825\n",
      "Epoch 11/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0079 - acc: 0.9978 - val_loss: 0.0753 - val_acc: 0.9813\n",
      "Epoch 12/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0059 - acc: 0.9986 - val_loss: 0.0788 - val_acc: 0.9819\n",
      "Epoch 13/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0048 - acc: 0.9987 - val_loss: 0.0853 - val_acc: 0.9827\n",
      "Epoch 14/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0039 - acc: 0.9989 - val_loss: 0.0872 - val_acc: 0.9818\n",
      "Epoch 15/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0031 - acc: 0.9992 - val_loss: 0.0838 - val_acc: 0.9838\n",
      "Epoch 16/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0025 - acc: 0.9994 - val_loss: 0.0890 - val_acc: 0.9836\n",
      "Epoch 17/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0020 - acc: 0.9996 - val_loss: 0.0872 - val_acc: 0.9845\n",
      "Epoch 18/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0017 - acc: 0.9996 - val_loss: 0.0887 - val_acc: 0.9827\n",
      "Epoch 19/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0014 - acc: 0.9996 - val_loss: 0.0956 - val_acc: 0.9830\n",
      "Epoch 20/20\n",
      "60000/60000 [==============================] - 3s - loss: 0.0013 - acc: 0.9997 - val_loss: 0.0971 - val_acc: 0.9833\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x11ba9de10>"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nw2fqUeCejcAAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11bb3b750>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "epochs = 20\n",
    "\n",
    "for activation in [None, 'sigmoid', 'tanh', 'relu']:\n",
    "    model = Sequential()\n",
    "    model.add(Dense(512, activation=activation, input_shape=(784,)))\n",
    "    model.add(Dense(num_classes, activation='softmax'))\n",
    "    model.compile(loss='categorical_crossentropy',\n",
    "              optimizer=RMSprop(),\n",
    "              metrics=['accuracy'])\n",
    "    history = model.fit(x_train, y_train,\n",
    "                    batch_size=batch_size,\n",
    "                    epochs=epochs,\n",
    "                    verbose=1,\n",
    "                    validation_data=(x_test, y_test))\n",
    "    \n",
    "    plt.plot(history.history['val_acc'])\n",
    "    \n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['None', 'sigmoid', 'tanh', 'relu'], loc='upper left')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python (hiredscore)",
   "language": "python",
   "name": "hiredscore"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}