convolutional_train.py

import tensorflow as tf
import input_data
import matplotlib.pyplot as plt
import matplotlib.cm as cm
import matplotlib

# Import training and testing data sets
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)

# Helper functions for initializing variables
def weight_variable(shape):
	initial = tf.truncated_normal(shape, stddev=0.1)
	return tf.Variable(initial)

def bias_variable(shape):
	initial = tf.constant(0.1, shape=shape)
	return tf.Variable(initial)

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

def max_pool_2x2(x):
	return tf.nn.max_pool(x, ksize=[1, 2, 2, 1], 
		strides=[1, 2, 2, 1], padding='SAME')

# Build the model

# Inputs and outputs
x = tf.placeholder(tf.float32, [None, 784])
y_ = tf.placeholder(tf.float32, [None, 10])

# First convolutional layer
# Patch weights and biases
W_conv1 = weight_variable([5,5,1,32])
b_conv1 = bias_variable([32])

# Reshape image for convolution
x_image = tf.reshape(x,[-1,28,28,1])

# Convolve and down-sample using max-pooling
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
h_pool1 = max_pool_2x2(h_conv1)

# Second convolutional layer
# Patch weights and biases
W_conv2 = weight_variable([5, 5, 32, 64])
b_conv2 = bias_variable([64])

# Convolve and down-sample using max-pooling
h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2)
h_pool2 = max_pool_2x2(h_conv2)

# Fully 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.nn.relu(tf.matmul(h_pool2_flat, w_fc1) + b_fc1)

# Dropout to avoid overfitting
keep_prob = tf.placeholder(tf.float32)
h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)

# Output layer of 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)

# Minimize cross entropy
cross_entropy = -tf.reduce_sum(y_*tf.log(y_conv))
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)

# Initialize session and variables
sess = tf.Session()
sess.run(tf.initialize_all_variables())

# For saving model state for later use
saver = tf.train.Saver()

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

# Train network, saving every 1000 steps and logging performance
for i in range(20000):
	batch = mnist.train.next_batch(50)
	if i%1000 == 0:
		print(sess.run(accuracy, feed_dict={x:batch[0], y_:batch[1], keep_prob: 1.0}))
		save_path = saver.save(sess, "MNIST_data/model.ckpt", global_step=i)
		print("Model checkpoint saved in file: %s" % save_path)
	sess.run(train_step, feed_dict={x: batch[0], y_:batch[1], keep_prob: 0.5})

# Test accuracy
print(sess.run(accuracy, feed_dict={x:mnist.test.images, y_:mnist.test.labels, keep_prob: 1.0}))