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

keep_prob=tf.placeholder(tf.float32)

def evaluation(logits, labels):
  correct = tf.nn.in_top_k(logits, labels, 1)
  return tf.reduce_sum(tf.cast(correct, tf.int32))

mnist=input_data.read_data_sets('MNIST_data',one_hot=True)
xs=tf.placeholder(tf.float32,[None,784])
ys=tf.placeholder(tf.float32,[None,10])

x_image = tf.reshape(xs,[-1,28,28,1])


#===================================================
def weight_v(shape):
    initial = tf.truncated_normal(shape,stddev=0.1)
    return tf.Variable(initial)

def bias_v(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')
#----------------------L1--------------------------
W_conv1 = weight_v([5,5,1,64])
b_conv1 = bias_v([64])

hid_conv1 = tf.nn.relu(conv2d(x_image,W_conv1)+b_conv1)
hid_pool1 = max_pool_2x2(hid_conv1)
#----------------------L2---------------------------
W_conv2 = weight_v([5,5,64,128])
b_conv2 = bias_v([128])

hid_conv2 = tf.nn.relu(conv2d(hid_pool1,W_conv2)+b_conv2)
hid_pool2 = max_pool_2x2(hid_conv2)
#---------------------------------------------------
W_fc1 = weight_v([7*7*128,1024])
b_fc1 = bias_v([1024])

hid_pool2flat = tf.reshape(hid_pool2,[-1,7*7*128])
hid_fc1 = tf.nn.relu(tf.matmul(hid_pool2flat,W_fc1)+b_fc1)
hid_fc1_dropout = tf.nn.dropout(hid_fc1,keep_prob=1.0)

#---------------------------------------------------
W_fc2 = weight_v([1024,10])
b_fc2 = bias_v([10])
prediction = tf.nn.softmax(tf.matmul(hid_fc1_dropout,W_fc2)+b_fc2)
#=================================================================
cross_entropy=tf.reduce_mean(-tf.reduce_sum(ys*tf.log(prediction),reduction_indices=[1]))
#loss
train=tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
#saver = tf.train.Saver()
#===================================================
def accuracy(v_xs,v_ys):
    global prediction
    y_pre=sess.run(prediction,feed_dict={xs:v_xs,keep_prob:1.0})
    judge=tf.equal(tf.argmax(y_pre,1),tf.argmax(v_ys,1))
    acc=tf.reduce_mean(tf.cast(judge,tf.float32))
    result=sess.run(acc,feed_dict={xs:v_xs,ys:v_ys,keep_prob:1.0})
    return result
#===================================================


with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    for i in range(0,1000):
        batch_xs,batch_ys=mnist.train.next_batch(50)
        sess.run(train,feed_dict={xs:batch_xs,ys:batch_ys,keep_prob:0.5})
        if i%20==0:
            print(i,accuracy(mnist.test.images[:1000],mnist.test.labels[:1000]))
#    save_path = saver.save(sess,"/home/ky/test/model/cnn.ckpt")