TensorFlow實戰3——TensorFlow實現CNN

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

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

def weight_variable(shape):
    '''初始化權重函數,truncated_normal創建標准差為0.2的截斷正態函數'''
    initial = tf.truncated_normal(shape, stddev=0.1)
    return tf.Variable(initial)

def bias_variable(shape):
    '''初始化偏置函數，由於使用ReLU要加一些正值0.1，避免死亡節點（dead neurons）'''
    initial = tf.constant(0.1, shape = shape)
    return tf.Variable(initial)

def conv2d(x, W):
    '''x:輸入 w:卷積參數 例[5, 5, 1, 32]:5, 5為卷積核尺寸
    1:為多少channel 彩色是3 灰度是1
    32:為卷積核的數量（這個卷積層會提取的多少類特征）
    strides:卷積模板移動的步長
    padding：代表邊界處理方式，SAME即輸入和輸出保持同樣尺寸'''
    return tf.nn.conv2d(x, W, strides = [1, 1, 1, 1], padding = 'SAME')

def max_pool_2x2(x):
    '''池化層函數 max_pool:最大池化函數'''
    return tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1],
                          padding='SAME')

#x:特征
x = tf.placeholder(tf.float32, [None, 784])
#y_真實的label
y_ = tf.placeholder(tf.float32, [None, 10])
'''卷積神經網絡會利用到原有的空間結構信息，因此需要將1D的輸入向量轉化為2D圖片結構（1x784->28x28）
因為只有一個顏色通道，故最終尺寸為[-1, 28, 28, 1]其中：-1代表樣本數量不固定，1代表顏色通道數量'''
x_image = tf.reshape(x, [-1, 28, 28, 1])

'''先定義weights和bias,然后使用conv2d函數進行卷積操作並加上偏置，
接着使用ReLU激活函數進行非線性處理，最好使用max_pool_2x2對卷積的輸出結果進行池化操作'''
w_conv1 = weight_variable([5, 5, 1, 32])
b_conv1 = bias_variable([32])
h_conv1 = tf.nn.relu(conv2d(x_image, w_conv1)+b_conv1)
h_pool1 = max_pool_2x2(h_conv1)

#定義第二個卷積層,不同在於特征變為64
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)
h_pool2 = max_pool_2x2(h_conv2)

'''經歷兩次2x2步長的最大池化，邊長變為1/4，圖片尺寸由28x28->7x7
由於第二個卷積層的卷積核數量為64，其輸出tensor尺寸為7x7x64。
使用tf.reshape函數對其變形，轉化為1D向量，然后連接一個全連接層，
隱含節點為1024，並使用Relu激活函數'''
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)

#為減輕過擬合，使用一個dropout層
keep_prob = tf.placeholder(tf.float32)
h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)

#dropout層輸出連softmax層，得到最后的概率輸出
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)

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

train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)

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

tf.global_variables_initializer().run()
for i in range(20000):
    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, train 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}))

step 0, train accuracy 0.22
step 100, train accuracy 0.9
step 200, train accuracy 0.88
step 300, train accuracy 0.9
step 400, train accuracy 0.96
step 500, train accuracy 0.96
step 600, train accuracy 0.98
step 700, train accuracy 0.96
step 800, train accuracy 0.98
step 900, train accuracy 0.96
step 1000, train accuracy 0.98
step 18000, train accuracy 1
step 18100, train accuracy 1
step 18200, train accuracy 0.98
step 18300, train accuracy 1
step 18400, train accuracy 1
step 18500, train accuracy 1
step 18600, train accuracy 1
step 18700, train accuracy 1
step 18800, train accuracy 1
step 18900, train accuracy 1
step 19000, train accuracy 1
step 19100, train accuracy 1
step 19200, train accuracy 1
step 19300, train accuracy 1
step 19400, train accuracy 1
step 19500, train accuracy 1
step 19600, train accuracy 1
step 19700, train accuracy 1
step 19800, train accuracy 1
step 19900, train accuracy 1
step 20000, train accuracy 1