mnist-advanced.py

## Tutorial >> https://www.tensorflow.org/get_started/mnist/beginners

import os
os.environ['TF_CPP_MIN_LOG_LEVEL']='2'
## https://github.com/tensorflow/tensorflow/issues/7778

import tensorflow as tf


from tensorflow.examples.tutorials.mnist import input_data


## --------- Utils --------- ##
# One should generally initialize weights with a small amount of noise for symmetry breaking, and to prevent 0 gradients.
def weight_variable(shape):
	initial = tf.truncated_normal(shape, stddev=0.1)
	return tf.Variable(initial)

# Since we're using ReLU neurons, it is also good practice to initialize them with a slightly positive initial bias to avoid "dead neurons"
def bias_variable(shape):
	initial = tf.constant(0.1, shape=shape)
	return tf.Variable(initial)

# Our convolutions uses a stride of one and are zero padded so that the output is the same size as the input.
def conv2d(x, W):
	return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME')

# Our pooling is plain old max pooling over 2x2 blocks.
def max_pool_2x2(x):
	return tf.nn.max_pool(x, ksize=[1, 2, 2, 1],
		strides=[1, 2, 2, 1], padding='SAME')
## --------- /Utils --------- ##



## load and extract the training and test data
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)

# mnist.train.images 55000 images, 784 pixels (28x28, but flatten as array ofr 784)
# mnist.train.labels 55000 labels, 10 one_hot dimentiona (0..9)

# Input, dimension number of training set (i.e. 55000) per the 28x28 flatten 784 bit array
x = tf.placeholder(tf.float32, shape=[None, 784])

# Expected output for training (one_hot labels)
y_ = tf.placeholder(tf.float32, shape=[None, 10])

## --------- First Convolution --------- ##
# The convolution will compute 32 features for each 5x5 patch. Its weight tensor will have a shape of [5, 5, 1, 32]. 
# The first two dimensions are the patch size, the next is the number of input channels, and the last is the number of output channels
W_conv1 = weight_variable([5, 5, 1, 32])

# We will also have a bias vector with a component for each output channel.
b_conv1 = bias_variable([32])

# To apply the layer, we first reshape x to a 4d tensor, with the second and third dimensions corresponding to image width and height, 
# and the final dimension corresponding to the number of color channels.
x_image = tf.reshape(x, [-1, 28, 28, 1])

# We then convolve x_image with the weight tensor, add the bias, apply the ReLU function, and finally max pool. 
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
# The max_pool_2x2 method will reduce the image size to 14x14.
h_pool1 = max_pool_2x2(h_conv1)

## --------- /First Convolution --------- ##


## --------- Second Convolution --------- ##

# The second layer will have 64 features for each 5x5 patch.
# The third dimension match the 32 output of conv1
W_conv2 = weight_variable([5, 5, 32, 64])
b_conv2 = bias_variable([64])


# Note the "x" os the output of convolution 1
h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2)

# Reduce to 7x7
h_pool2 = max_pool_2x2(h_conv2)

## --------- /Second Convolution --------- ##

## --------- Densely Connected --------- ##

# Now that the image size has been reduced to 7x7, we add a fully-connected layer with 1024 neurons to allow processing on the entire image. 
# We reshape the tensor from the pooling layer into a batch of vectors, multiply by a weight matrix, add a bias, and apply a ReLU.
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)

## --------- /Densely Connected --------- ##

## --------- Dropout --------- ##

# To reduce overfitting, we will apply dropout before the readout layer. We create a placeholder for the probability that a neuron's output 
# is kept during dropout. This allows us to turn dropout on during training, and turn it off during testing. 
# TensorFlow's tf.nn.dropout op automatically handles scaling neuron outputs in addition to masking them, so dropout just works without any additional scaling
keep_prob = tf.placeholder(tf.float32)
h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)

## --------- /Dropout --------- ##

## --------- Readout Layer --------- ##

# Finally, we add a layer, just like for the one layer softmax regression above.
W_fc2 = weight_variable([1024, 10])
b_fc2 = bias_variable([10])

# The Final Model
y_conv = tf.matmul(h_fc1_drop, W_fc2) + b_fc2

## --------- /Readout Layer --------- ##

# Build the loss
cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y_, logits=y_conv))

# train step with the AdamOptimizer replacing the gradient decent
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)

# Compute prediction accuracy (this is for logging the progress)
correct_prediction = tf.equal(tf.argmax(y_conv, 1), tf.argmax(y_, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))


with tf.Session() as sess:

	# Initialize
	sess.run(tf.global_variables_initializer())

	# 20k training operations
	for i in range(20000):
		# get the batch
		batch = mnist.train.next_batch(50)

		# print the learning progress every 100 iteration
		if i % 100 == 0:
			train_accuracy = accuracy.eval(feed_dict={
					x: batch[0], y_: batch[1], keep_prob: 1.0})
			print('step %d, training accuracy %g' % (i, train_accuracy))

		# train (this will do the back propagation)
		# Note: We will include the additional parameter keep_prob in feed_dict to control the dropout rate.
		train_step.run(feed_dict={x: batch[0], y_: batch[1], keep_prob: 0.5})

	# Now training is done, evaluate the test images
	print('test accuracy %g' % accuracy.eval(feed_dict={
			x: mnist.test.images, y_: mnist.test.labels, keep_prob: 1.0}))