tensorflow 官网教程 - Deep MNIST for Experts - 代码及注解

import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
import tensorflow as tf

# Import dataset for experiment
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets('MNIST_data', one_hot=True)


def weight_variable(shape): # weight initialization
    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')


sess = tf.InteractiveSession()
x = tf.placeholder(tf.float32, [None, 784]) # None means a dimension can be of any length
y_ = tf.placeholder(tf.float32, [None, 10])


# First Convolutional Layer
W_conv1 = weight_variable([5,5,1,32])
    # heights x widths x channels x number of filters
b_conv1 = bias_variable([32])
    # 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. We will also have a bias vector with a component for each output channel.
x_image = tf.reshape(x, [-1,28,28,1])
    # segmenting the image A: and then 1 image was segmented to several images
    # number of images(1) x heights(28) x widths(28) x channels(1)
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
    # size of each image is [28,28] with 32 channels now
h_pool1 = max_pool_2x2(h_conv1)
    # size of each image is [14,14] with 32 channels now


# Second Convolutional Layer
W_conv2 = weight_variable([5,5,32,64])
b_conv2 = bias_variable([64])
h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2)
    # size of each image is [14,14] with 64 channels now
h_pool2 = max_pool_2x2(h_conv2)
    # size of each image is [7,7] with 64 channels now


# 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.nn.relu(tf.matmul(h_pool2_flat, W_fc1) + b_fc1)


# 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)


# Readout Layer
W_fc2 = weight_variable([1024, 10])
b_fc2 = bias_variable([10])
y_conv = tf.matmul(h_fc1_drop, W_fc2) + b_fc2


# Train and Evaluate the Model
cross_entropy = tf.reduce_mean(
    tf.nn.softmax_cross_entropy_with_logits(labels=y_, logits=y_conv))

train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
correct_prediction = tf.equal(tf.argmax(y_,1),tf.argmax(y_conv,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

sess.run(tf.global_variables_initializer())
for i in range(6000):
    batch = mnist.train.next_batch(50)
    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_step.run(feed_dict={x: batch[0], y_:batch[1], keep_prob: 0.5})

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

# ====================================
# Note here:
#   session: the connection between tensorflow and the C++ backend
#   common usage for Tensorflow:
#       - first create a graph
#       - then launch the graph in a session
#   weight initialization for CNN
#       - with a small amount of noise for symmetry breaking & to prevent 0 gradients
#       - when using ReLU neuron: initialize them with a slightly positive initial bias to avoid dead neurons