MNIST 数据分类

分类和回归的区别在于输出变量的类型上。通俗理解定量输出是回归，或者说是连续变量预测；定性输出是分类，

或者说是离散变量预测。如预测房价这是一个回归任务；把东西分成几类, 比如猫狗猪牛，就是一个分类任务。

 1 import tensorflow as tf
 2 import numpy as np
 3 from tensorflow.examples.tutorials.mnist import input_data   #首先准备数据
 4 mnist =input_data.read_data_sets('MNIST_data',one_hot=True)
   #调用add_layer函数搭建一个最简单的训练网络结构，只有输入层和输出层。  
 5 def add_layer(inputs, in_size, out_size,activation_function=None):
 6     Weights = tf.Variable(tf.random_normal([in_size, out_size]))
 7     biases = tf.Variable(tf.zeros([1, out_size]) + 0.1)
 8     Wx_plus_b = tf.matmul(inputs, Weights) + biases
 9     if activation_function is None:
10         outputs = Wx_plus_b
11     else:
12         outputs = activation_function(Wx_plus_b)
13     return outputs
14 def compute_accuracy(v_xs, v_ys):
15     global prediction
16     y_pre = sess.run(prediction, feed_dict={xs: v_xs})
17     correct_prediction = tf.equal(tf.argmax(y_pre,1), tf.argmax(v_ys,1))
18     accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
19     result = sess.run(accuracy, feed_dict={xs: v_xs, ys: v_ys})
20     return result
21 #其中输入数据是784个特征，输出数据是10个特征，激励采用softmax函数
22 xs=tf.placeholder(tf.float32,[None,784]) #28*28
23 ys=tf.placeholder(tf.float32,[None,10])
24 prediction=add_layer(xs,784,10,activation_function=tf.nn.softmax)
25 #loss
26 cross_entropy=tf.reduce_mean(-tf.reduce_sum(ys*tf.log(prediction),reduction_indices=[1]))
27 
28 train_step=tf.train.GradientDescentOptimizer(0.5).minimize(cross_entropy)
29 
30 init = tf.global_variables_initializer()
31 sess=tf.Session()
32 sess.run(init)
33 for i in range(1000):
34     batch_xs,batch_ys=mnist.train.next_batch(100)
35     sess.run(train_step,feed_dict={xs:batch_xs,ys:batch_ys})
36     if i%50==0:
37         print(compute_accuracy(
38             mnist.test.images,mnist.test.labels))

MNIST 数据 分类

MNIST 数据分类