mnist 训练 python

---恢复内容开始---

import numpy as np
import tensorflow as tf
import struct
import cv2
import input_data
import matplotlib.pyplot as plt
from sklearn.neighbors import KNeighborsClassifier
import numpy as np
import os

os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2 '

# #图像数据传入
# #原始数据 3维训练与测试数据和1维训练与测试标签
# data_or_tr=input_data.f_tr
# data_or_trla=input_data.f_tr_lab
# data_or_sa=input_data.f_tr
# data_or_sala=input_data.f_tr_lab
#
#
# #已将原始数据转为2维训练与测试数据和1维训练与测试标签
# data_tr=input_data.T_tr
# data_trla=input_data.T_tr_lab
# data_sa=input_data.T_tr
# data_sala=input_data.T_tr_lab
#
# #显示二维输入图像模块函数,只显示一张
# def showimage(imag):
# plt.subplot(111) #行和列在第1个画布
# plt.imshow(imag, cmap=plt.get_cmap('gray')) #
# plt.show()

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

#定义一个获取卷积核的函数
def weight_variable(shape):
initial = tf.truncated_normal(shape, stddev=0.1) # 产生正态分布，最终获得shape维度的数据
return tf.Variable(initial)
#定义一个获取偏置值的函数
def bias_variable(shape):
initial = tf.constant(0.1,shape=shape) #初始化偏置量为0.1 最终获得shape维度的数据
return tf.Variable(initial)
#定义一个卷积函数
def conv2d(x,W):
return tf.nn.conv2d(x,W,[1,1,1,1],padding="SAME") # x为输入图像的张量 [1,1,1,1]表示步长，W为滤波数组
#定义一个池化函数
def max_pool_2x2(x):
return tf.nn.max_pool(x,ksize=[1,2,2,1], strides=[1,2,2,1],padding="VALID")

if __name__ == "__main__":
mnist = input_data.read_data_sets("C:\\Users\\HHQ\Desktop\\tangjun\\minist\\minist_data\\minist",one_hot=True)
x = tf.placeholder(shape=[None,28*28],dtype=tf.float32) #定义x为二维张量的占位符
lable = tf.placeholder(shape=[None,10],dtype=tf.float32) #
x_image = tf.reshape(x,[-1,28,28,1]) # -1 表示不定
#第一个卷积层
W_conv1 = weight_variable([5,5,1,32]) #正太分布产生卷积核位为啥是32呢？
b_conv1 = bias_variable([32]) #产生32个片质量
h_conv1 = tf.nn.relu(conv2d(x_image,W_conv1)+b_conv1) #我觉得应该产生5*5的数组
h_pool1 = max_pool_2x2(h_conv1) #
#14*14*32
#第二个卷积层
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)
#7*7*64
#全连接层，输出为1024维向量
W_fc1 = weight_variable([7*7*64,1024]) # 正太分布随机产生变量 variable变量
b_fc1 = weight_variable([1024]) # 正太分布随机产生变量 variable变量
h_pool2_flat = tf.reshape(h_pool2,[-1,7*7*64]) #h_pool2_flat是一个二维的数据
h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat,W_fc1)+b_fc1) #矩阵相乘，二层卷积池化后产生举证与随机生成的举证相乘
keep_prob = tf.placeholder(tf.float32) #未限制形状
h_fc1_dropout = tf.nn.dropout(h_fc1,keep_prob=keep_prob)

#把1024维向量转换成10维，对应10个类别
W_fc2 = weight_variable([1024,10])
b_fc2 = weight_variable([10])
y_conv = tf.matmul(h_fc1,W_fc2)+b_fc2

#直接使用tf.nn.softmax_cross_entropy_with_logits直接计算交叉熵
cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=lable,logits=y_conv))
#定义train_step
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
#定义测试的准确率
correct_prediction = tf.equal(tf.argmax(y_conv,1),tf.argmax(lable,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))
# 创建Session和变量初始化
sess = tf.InteractiveSession()
sess.run(tf.global_variables_initializer())
#训练20000步
for i in range(200):
batch = mnist.train.next_batch(50)
if i % 10==0:
train_accuracy = sess.run(accuracy,feed_dict={
x:batch[0],lable:batch[1],keep_prob: 1.0})
print("step %d, training accuracy %g" % (i, train_accuracy))

_ = sess.run(train_step, feed_dict={x: batch[0], lable: batch[1], keep_prob: 0.5})

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

---恢复内容结束---