from tensorflow.examples.tutorials.mnist import input_data 
import tensorflow as tf 
mnist = input_data.read_data_sets("MNSIT_data/", one_hot=True) 
sess = tf.InteractiveSession() 
 
 
# In[2]: 
#由于W和b在各層中均要用到，先定義乘函數(shù)。 
#tf.truncated_normal：截斷正態(tài)分布，即限制范圍的正態(tài)分布 
def weight_variable(shape): 
  initial = tf.truncated_normal(shape, stddev=0.1) 
  return tf.Variable(initial) 
 
 
# In[7]: 
#bias初始化值0.1. 
def bias_variable(shape): 
  initial = tf.constant(0.1, shape=shape) 
  return tf.Variable(initial) 
 
 
# In[12]: 
#tf.nn.conv2d:二維的卷積 
#conv2d(input, filter, strides, padding, use_cudnn_on_gpu=None,data_format=None, name=None) 
#filter:A 4-D tensor of shape 
#   `[filter_height, filter_width, in_channels, out_channels]` 
#strides:步長，都是1表示所有點都不會被遺漏。1-D 4值，表示每歌dim的移動步長。 
# padding:邊界的處理方式，“SAME"、"VALID”可選 
def conv2d(x, W): 
  return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME') 
 
#tf.nn.max_pool:最大值池化函數(shù)，即求2*2區(qū)域的最大值，保留最顯著的特征。 
#max_pool(value, ksize, strides, padding, data_format="NHWC", name=None) 
#ksize:池化窗口的尺寸 
#strides:[1,2,2,1]表示橫豎方向步長為2 
def max_pool_2x2(x): 
  return tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides = [1, 2, 2, 1], padding='SAME') 
 
 
x = tf.placeholder(tf.float32, [None, 784]) 
y_ = tf.placeholder(tf.float32, [None, 10]) 
#tf.reshape:tensor的變形函數(shù)。 
#-1：樣本數(shù)量不固定 
#28,28：新形狀的shape 
#1:顏色通道數(shù) 
x_image = tf.reshape(x, [-1, 28, 28, 1]) 
 
 
#卷積層包含三部分：卷積計算、激活、池化 
#[5,5,1,32]表示卷積核的尺寸為5×5, 顏色通道為1, 有32個卷積核 
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) 
 
 
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(jīng)過2次2×2的池化后，圖像的尺寸變?yōu)?×7,第二個卷積層有64個卷積核，生成64類特征，因此，卷積最后輸出為7×7×64. 
#tensor進入全連接層之前，先將64張二維圖像變形為1維圖像，便于計算。 
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) 
 
 
#對全連接層做dropot 
keep_prob = tf.placeholder(tf.float32) 
h_fc1_dropout = tf.nn.dropout(h_fc1, keep_prob) 
 
 
#又一個全連接后foftmax分類 
W_fc2 = weight_variable([1024, 10]) 
b_fc2 = bias_variable([10]) 
y_conv = tf.nn.softmax(tf.matmul(h_fc1_dropout, W_fc2) + b_fc2) 
 
 
cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_*tf.log(y_conv), reduction_indices=[1])) 
#AdamOptimizer：Adam優(yōu)化函數(shù) 
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)) 
 
 
#訓練，并且每100個batch計算一次精度 
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, training accuracy %g" %(i, train_accuracy)) 
  train_step.run(feed_dict={x:batch[0], y_:batch[1], keep_prob:0.5}) 
 
 
#在測試集上測試 
print("test accuracy %g"%accuracy.eval(feed_dict={x:mnist.test.images, y_:mnist.test.labels, keep_prob:1.0}))