ConvNet.py

import os
import time

# Load MNIST dataset
from tensorflow.examples.tutorials.mnist import input_data

mnist = input_data.read_data_sets('MNIST_data', one_hot=True)

# Import Tensorflow and start a session
import tensorflow as tf

sess = tf.InteractiveSession()


def create_summary(var, name):
    mean = tf.reduce_mean(var)
    tf.scalar_summary('mean/' + name, mean)
    stddev = tf.sqrt(tf.reduce_sum(tf.square(var - mean)))
    tf.scalar_summary('std_dev/' + name, tf.sqrt(tf.reduce_sum(tf.square(var - mean))))
    tf.scalar_summary('max/' + name, tf.reduce_max(var))
    tf.scalar_summary('min/' + name, tf.reduce_min(var))
    tf.histogram_summary('y/' + name, var)
    return None


def weight_variable(shape):
    '''
    Initialize weights
    :param shape: shape of weights, e.g. [w, h ,Cin, Cout] where
    w: width of the filters
    h: height of the filters
    Cin: the number of the channels of the filters
    Cout: the number of filters
    :return: a tensor variable for weights with initial values
    '''

    # IMPLEMENT YOUR WEIGHT_VARIABLE HERE
    initial = tf.random_normal(shape, stddev=0.1)
    # initial = tf.get_variable("Weights_1/", shape=shape, initializer=tf.contrib.layers.xavier_initializer())
    W = tf.Variable(initial)
    return W


def bias_variable(shape):
    '''
    Initialize biases
    :param shape: shape of biases, e.g. [Cout] where
    Cout: the number of filters
    :return: a tensor variable for biases with initial values
    '''

    # IMPLEMENT YOUR BIAS_VARIABLE HERE
    # initial = tf.constant(0.1, shape=shape)
    initial = tf.random_normal(shape, stddev=0.1)
    # initial = tf.get_variable("Weights_bias/", shape=shape,initializer=tf.contrib.layers.xavier_initializer())
    b = tf.Variable(initial)
    return b


def conv2d(x, W):
    '''
    Perform 2-D convolution
    :param x: input tensor of size [N, W, H, Cin] where
    N: the number of images
    W: width of images
    H: height of images
    Cin: the number of channels of images
    :param W: weight tensor [w, h, Cin, Cout]
    w: width of the filters
    h: height of the filters
    Cin: the number of the channels of the filters = the number of channels of images
    Cout: the number of filters
    :return: a tensor of features extracted by the filters, a.k.a. the results after convolution
    '''

    # IMPLEMENT YOUR CONV2D HERE

    h_conv = tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME')
    return h_conv


def max_pool_2x2(x):
    '''
    Perform non-overlapping 2-D maxpooling on 2x2 regions in the input data
    :param x: input data
    :return: the results of maxpooling (max-marginalized + downsampling)
    '''

    # IMPLEMENT YOUR MAX_POOL_2X2 HERE
    h_max = tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
    return h_max


def main():
    # Specify training parameters
    result_dir = './results/'  # directory where the results from the training are saved
    max_step = 5500  # the maximum iterations. After max_step iterations, the training will stop no matter what

    start_time = time.time()  # start timing

    # FILL IN THE CODE BELOW TO BUILD YOUR NETWORK

    # placeholders for input data and input labeles
    x = tf.placeholder(tf.float32, [None, 784])
    y_ = tf.placeholder(tf.float32, [None, 10])

    # reshape the input image
    x_image = tf.reshape(x, [-1, 28, 28, 1])

    # first convolutional layer
    W_conv1 = weight_variable([5, 5, 1, 32])
    b_conv1 = bias_variable([32])
    h_conv1 = tf.sigmoid(conv2d(x_image, W_conv1) + b_conv1)
    h_pool1 = max_pool_2x2(h_conv1)

    # second convolutional layer
    W_conv2 = weight_variable([5, 5, 32, 64])
    b_conv2 = bias_variable([64])
    h_conv2 = tf.sigmoid(conv2d(h_pool1, W_conv2) + b_conv2)
    h_pool2 = max_pool_2x2(h_conv2)

    # densely connected layer
    W_fc1 = weight_variable([7 * 7 * 64, 1024])
    b_fc1 = bias_variable([1024])
    h_pool2_flat = tf.reshape(h_pool2, [-1, 7 * 7 * 64])
    h_fc1 = tf.sigmoid(tf.matmul(h_pool2_flat, W_fc1) + b_fc1)

    # dropout
    keep_prob = tf.placeholder(tf.float32)
    h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)

    # softmax
    W_fc2 = weight_variable([1024, 10])
    b_fc2 = bias_variable([10])

    y_conv = tf.nn.softmax(tf.matmul(h_fc1_drop, W_fc2) + b_fc2)

    # FILL IN THE FOLLOWING CODE TO SET UP THE TRAINING

    # setup training
    cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(y_conv), reduction_indices=[1]))
    train_step = tf.train.RMSPropOptimizer(1e-4).minimize(cross_entropy)

    var = -tf.reduce_sum(y_ * tf.log(y_conv), reduction_indices=[1])

    create_summary(var, "Training_Loss")

    create_summary(W_conv1, "Weight_conv1")
    create_summary(b_conv1, "bias_conv1")
    create_summary(h_conv1, "h_conv1")
    create_summary(h_pool1, "h_pool1")
    create_summary(W_conv2, "Weight_conv2")
    create_summary(b_conv2, "bias_conv2")
    create_summary(h_conv2, "h_conv2")
    create_summary(h_pool2, "h_pool2")
    create_summary(W_fc1, "Weight_fc1")
    create_summary(b_fc1, "bias_fc1")
    create_summary(h_fc1, "activation_fc1")
    create_summary(W_fc2, "Weight_fc2")
    create_summary(b_fc2, "bias_fc2")
    create_summary(y_conv, "activation_fc2")

    correct_prediction = tf.equal(tf.argmax(y_conv, 1), tf.argmax(y_, 1))
    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

    # Add a scalar summary for the snapshot loss.
    tf.scalar_summary(cross_entropy.op.name, cross_entropy)
    # Build the summary operation based on the TF collection of Summaries.
    summary_op = tf.merge_all_summaries()

    # Add the variable initializer Op.
    init = tf.initialize_all_variables()

    # Create a saver for writing training checkpoints.
    saver = tf.train.Saver()

    # Instantiate a SummaryWriter to output summaries and the Graph.
    summary_writer = tf.train.SummaryWriter(result_dir, sess.graph)

    # Run the Op to initialize the variables.
    sess.run(init)

    # run the training
    for i in range(max_step + 1):
        batch = mnist.train.next_batch(50)  # make the data batch, which is used in the training iteration.
        # the batch size is 50
        if i % 100 == 0:
            # output the training accuracy every 100 iterations
            train_accuracy = accuracy.eval(feed_dict={
                x: batch[0], y_: batch[1], keep_prob: 1.0})

            test_accuracy = accuracy.eval(feed_dict={
                x: mnist.test.images, y_: mnist.test.labels, keep_prob: 1.0})

            validation_accuracy = accuracy.eval(feed_dict={
                x: mnist.validation.images, y_: mnist.validation.labels, keep_prob: 1.0})

            # Update the events file which is used to monitor the training (in this case,
            # only the training loss is monitored)
            summary_str = sess.run(summary_op, feed_dict={x: batch[0], y_: batch[1], keep_prob: 0.5})
            summary_writer.add_summary(summary_str, i)
            summary_writer.flush()

        # save the checkpoints every 1100 iterations
        if i % 1100 == 0 or i == max_step:
            print("step %d, training accuracy %g, test accuracy %g, validation accuracy %g " % (
            i, train_accuracy, test_accuracy, validation_accuracy))

            checkpoint_file = os.path.join(result_dir, 'checkpoint')
            saver.save(sess, checkpoint_file, global_step=i)

        train_step.run(feed_dict={x: batch[0], y_: batch[1], keep_prob: 0.5})  # run one train_step

    # print test error
    print("test accuracy %g" % accuracy.eval(feed_dict={
        x: mnist.test.images, y_: mnist.test.labels, keep_prob: 1.0}))

    stop_time = time.time()
    print('The training takes %f second to finish' % (stop_time - start_time))


if __name__ == "__main__":
    main()


#
# Ref:
# 1) https://www.tensorflow.org
# 2) https://ankitlab.co/