notebooks/intermediate_net_in_tensorflow.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Intermediate Neural Network in TensorFlow"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this notebook, we improve on our [introductory shallow net](https://github.com/jonkrohn/DLTFpT/blob/master/notebooks/shallow_net_in_tensorflow.ipynb) by incorporating the theory we've covered since."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/jonkrohn/DLTFpT/blob/master/notebooks/intermediate_net_in_tensorflow.ipynb)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Load dependencies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "import tensorflow\n",
    "from tensorflow.keras.datasets import mnist\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense\n",
    "from tensorflow.keras.optimizers import SGD\n",
    "from matplotlib import pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Load data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "(X_train, y_train), (X_valid, y_valid) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Preprocess data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = X_train.reshape(60000, 784).astype('float32')\n",
    "X_valid = X_valid.reshape(10000, 784).astype('float32')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train /= 255\n",
    "X_valid /= 255"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "n_classes = 10\n",
    "y_train = to_categorical(y_train, n_classes)\n",
    "y_valid = to_categorical(y_valid, n_classes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Design neural network architecture"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "model.add(Dense(64, activation='relu', input_shape=(784,)))\n",
    "model.add(Dense(64, activation='relu'))\n",
    "model.add(Dense(10, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense (Dense)                (None, 64)                50240     \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 64)                4160      \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                650       \n",
      "=================================================================\n",
      "Total params: 55,050\n",
      "Trainable params: 55,050\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "4096"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(64*64)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "4160"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(64*64)+64"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Configure model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# N.B.: learning rate is order of magnitude quicker relative to shallow net\n",
    "model.compile(loss='categorical_crossentropy', optimizer=SGD(lr=0.1), metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Train!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/20\n",
      "60000/60000 [==============================] - 1s 15us/sample - loss: 0.4824 - accuracy: 0.8625 - val_loss: 0.2545 - val_accuracy: 0.9262\n",
      "Epoch 2/20\n",
      "60000/60000 [==============================] - 1s 20us/sample - loss: 0.2318 - accuracy: 0.9316 - val_loss: 0.1963 - val_accuracy: 0.9426\n",
      "Epoch 3/20\n",
      "60000/60000 [==============================] - 1s 20us/sample - loss: 0.1778 - accuracy: 0.9480 - val_loss: 0.1525 - val_accuracy: 0.9558\n",
      "Epoch 4/20\n",
      "60000/60000 [==============================] - 1s 21us/sample - loss: 0.1452 - accuracy: 0.9583 - val_loss: 0.1330 - val_accuracy: 0.9601\n",
      "Epoch 5/20\n",
      "60000/60000 [==============================] - 1s 18us/sample - loss: 0.1237 - accuracy: 0.9639 - val_loss: 0.1219 - val_accuracy: 0.9630\n",
      "Epoch 6/20\n",
      "60000/60000 [==============================] - 1s 19us/sample - loss: 0.1077 - accuracy: 0.9686 - val_loss: 0.1071 - val_accuracy: 0.9674\n",
      "Epoch 7/20\n",
      "60000/60000 [==============================] - 1s 18us/sample - loss: 0.0953 - accuracy: 0.9718 - val_loss: 0.0989 - val_accuracy: 0.9701\n",
      "Epoch 8/20\n",
      "60000/60000 [==============================] - 1s 20us/sample - loss: 0.0857 - accuracy: 0.9743 - val_loss: 0.1000 - val_accuracy: 0.9696\n",
      "Epoch 9/20\n",
      "60000/60000 [==============================] - 1s 21us/sample - loss: 0.0767 - accuracy: 0.9778 - val_loss: 0.0940 - val_accuracy: 0.9711\n",
      "Epoch 10/20\n",
      "60000/60000 [==============================] - 1s 20us/sample - loss: 0.0707 - accuracy: 0.9796 - val_loss: 0.0876 - val_accuracy: 0.9732\n",
      "Epoch 11/20\n",
      "60000/60000 [==============================] - 1s 21us/sample - loss: 0.0651 - accuracy: 0.9808 - val_loss: 0.0842 - val_accuracy: 0.9746\n",
      "Epoch 12/20\n",
      "60000/60000 [==============================] - 1s 21us/sample - loss: 0.0598 - accuracy: 0.9827 - val_loss: 0.0914 - val_accuracy: 0.9731\n",
      "Epoch 13/20\n",
      "60000/60000 [==============================] - 1s 18us/sample - loss: 0.0548 - accuracy: 0.9837 - val_loss: 0.0886 - val_accuracy: 0.9746\n",
      "Epoch 14/20\n",
      "60000/60000 [==============================] - 1s 19us/sample - loss: 0.0498 - accuracy: 0.9853 - val_loss: 0.0882 - val_accuracy: 0.9744\n",
      "Epoch 15/20\n",
      "60000/60000 [==============================] - 1s 22us/sample - loss: 0.0472 - accuracy: 0.9863 - val_loss: 0.0834 - val_accuracy: 0.9751\n",
      "Epoch 16/20\n",
      "60000/60000 [==============================] - 1s 17us/sample - loss: 0.0434 - accuracy: 0.9869 - val_loss: 0.0855 - val_accuracy: 0.9747\n",
      "Epoch 17/20\n",
      "60000/60000 [==============================] - 1s 19us/sample - loss: 0.0410 - accuracy: 0.9880 - val_loss: 0.0832 - val_accuracy: 0.9741\n",
      "Epoch 18/20\n",
      "60000/60000 [==============================] - 1s 19us/sample - loss: 0.0372 - accuracy: 0.9896 - val_loss: 0.0818 - val_accuracy: 0.9756\n",
      "Epoch 19/20\n",
      "60000/60000 [==============================] - 1s 16us/sample - loss: 0.0347 - accuracy: 0.9902 - val_loss: 0.0889 - val_accuracy: 0.9723\n",
      "Epoch 20/20\n",
      "60000/60000 [==============================] - 1s 17us/sample - loss: 0.0329 - accuracy: 0.9902 - val_loss: 0.0797 - val_accuracy: 0.9745\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.keras.callbacks.History at 0x7f3cb0286668>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# N.B.: number of epochs is order of magnitude smaller relative to shallow net\n",
    "model.fit(X_train, y_train, batch_size=128, epochs=20, verbose=1, validation_data=(X_valid, y_valid))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Performing inference"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "valid_0 = X_valid[0].reshape(1, 784)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[2.0512989e-07, 1.2484303e-06, 7.8251987e-04, 3.2485623e-02,\n",
       "        2.7430804e-09, 5.0794038e-06, 1.3936890e-12, 9.6642137e-01,\n",
       "        2.1935640e-04, 8.4524676e-05]], dtype=float32)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.predict(valid_0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([7])"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.predict_classes(valid_0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}