SimpleFeedForward.py

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data

mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)

learning_rate = 0.5
training_epochs = 1000
batch_size = 64

number_of_input = 784
number_of_output = 10

layer_1_size = 128
layer_2_size = 64

is_train = False

with tf.variable_scope("input"):
    X = tf.placeholder(tf.float32, shape=[None, number_of_input])

with tf.variable_scope("layer1"):
    weights = tf.get_variable(name="weights1", shape=[number_of_input, layer_1_size],
                              initializer=tf.contrib.layers.xavier_initializer())

    biases = tf.get_variable(name="biases1", shape=[layer_1_size], initializer=tf.zeros_initializer())
    layer_1_output = tf.nn.relu(tf.matmul(X, weights) + biases)

with tf.variable_scope("layer2"):
    weights = tf.get_variable(name="weights2", shape=[layer_1_size, layer_2_size],
                              initializer=tf.contrib.layers.xavier_initializer())

    biases = tf.get_variable(name="biases2", shape=[layer_2_size], initializer=tf.zeros_initializer())
    layer_2_output = tf.nn.relu(tf.matmul(layer_1_output, weights) + biases)

with tf.variable_scope("output"):
    weights = tf.get_variable(name="weightsOut", shape=[layer_2_size, number_of_output],
                              initializer=tf.contrib.layers.xavier_initializer())

    biases = tf.get_variable(name="biasesOut", shape=[number_of_output], initializer=tf.zeros_initializer())
    output = tf.nn.softmax(tf.matmul(layer_2_output, weights) + biases)

with tf.variable_scope("cost"):
    Y = tf.placeholder(tf.float32, shape=[None, number_of_output])
    cost = tf.reduce_mean(-tf.reduce_sum(Y * tf.log(output), reduction_indices=[1]))

with tf.variable_scope("accuracy"):
    correct_prediction = tf.equal(tf.argmax(output, 1), tf.argmax(Y, 1))
    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

with tf.variable_scope("train"):
    optimizer = tf.train.AdadeltaOptimizer(learning_rate).minimize(cost)

with tf.variable_scope("log"):
    tf.summary.scalar("Current_Cost", cost)
    tf.summary.scalar("Accuracy", accuracy)
    summary = tf.summary.merge_all()

saver = tf.train.Saver()
# Training Loop
with tf.Session() as session:
    session.run(tf.global_variables_initializer())

    training_writer = tf.summary.FileWriter('LogsFeedForward/training', session.graph)
    testing_writer = tf.summary.FileWriter('LogsFeedForward/testing', session.graph)
    accuracy_writer = tf.summary.FileWriter('LogsFeedForward/accuracy', session.graph)

    for epoch in range(training_epochs):
        train_batch_data, train_batch_labels = mnist.train.next_batch(batch_size)
        test_batch_data, test_batch_labels = mnist.test.next_batch(batch_size)

        session.run(optimizer, feed_dict={X: train_batch_data, Y: train_batch_labels})
        if epoch % 5 == 0:
            training_cost, training_summary = session.run([cost, summary], feed_dict={X: train_batch_data,
                                                                                      Y: train_batch_labels})
            testing_cost, testing_summary = session.run([cost, summary], feed_dict={X: test_batch_data,
                                                                                    Y: test_batch_labels})
            acc, accuracy_summary = session.run([accuracy, summary], feed_dict={X: mnist.test.images,
                                                                                Y: mnist.test.labels})

            # Save To Log File
            training_writer.add_summary(training_summary, epoch)
            testing_writer.add_summary(testing_summary, epoch)
            accuracy_writer.add_summary(accuracy_summary, epoch)

            print("Training Cost: ", training_cost, " Testing Cost: ", testing_cost, " Accuracy: ", acc)

    save_path = saver.save(session, 'saveFeedForward/trained_model.ckpt')
    print("Model Saved")

    final_accuracy = session.run(accuracy, feed_dict={X: mnist.test.images, Y: mnist.test.labels})

print("Training Complete with: ", final_accuracy, " accuracy.")