MNIST是在机器学习领域中的一个经典问题。为了学习机器学习和TensorFlow库的使用,使用TF构造一个softmax回归网络模型去识别手写数字。以下内容请参考TensorFlow中文社区(http://www.tensorfly.cn/tfdoc/tutorials/mnist_beginners.html)
1)MNIST介绍
2)softmax回归介绍
3)回归模型的训练和评估
具体代码如下:
"""
分类对象:MNIST手写数字识别
分类方法:使用一个2层NN,即一个非线性变换来识别
input ------------>output
784d softmax 10d
"""
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
from tensorflow.examples.tutorials.mnist import input_data
# loading mnist data
mnist = input_data.read_data_sets('MNIST_data',one_hot=True)
def add_layer(inputs,in_size,out_size,activation_function=None):
Weights = tf.Variable(tf.random_normal([in_size, out_size]), name='W')
biases = tf.Variable(tf.zeros([1, out_size]) + 0.1, name='b')
Wx_plus_b = tf.matmul(inputs, Weights) + biases
if activation_function is None: # 如何激活函数为空,则是线性函数
outputs = Wx_plus_b
else:
outputs = activation_function(Wx_plus_b)
return outputs
def compute_accuracy(vxs,vys):
global prediction
y_pre = sess.run(prediction,feed_dict={xs:vxs})
err = tf.equal(tf.argmax(y_pre,1),tf.argmax(vys,1))
acc = tf.reduce_mean(tf.cast(err,tf.float32))
result = sess.run(acc,feed_dict={xs:vxs,ys:vys})
return result
# define placeholder for inputs to network
xs = tf.placeholder(tf.float32,[None,784]) #28*28
ys = tf.placeholder(tf.float32,[None,10])
# add output layer
prediction = add_layer(xs,784,10,activation_function=tf.nn.softmax)
# the error between prediction and real data
loss = tf.reduce_mean(-tf.reduce_sum(ys*tf.log(prediction),
reduction_indices=[1]))
train_step = tf.train.GradientDescentOptimizer(0.5).minimize(loss)
#important step
sess = tf.Session()
sess.run(tf.global_variables_initializer()) #tf.initialize_all_variables在2017-03-02之后删除,改为global_variables_initializer
plt.figure()
plt.axes([0,10000,0,1])
plt.ion()
batch_size = 100
for i in range(10000):
batch_xs,batch_ys = mnist.train.next_batch(batch_size)
sess.run(train_step,feed_dict={xs:batch_xs,ys:batch_ys})
if i%50==0:
acc =compute_accuracy(mnist.test.images,
mnist.test.labels)
print(acc)
plt.plot(i,acc,'b.-')
plt.pause(0.00000001)
plt.ioff()
plt.show()