tensorflow分类任务MNIST数据集

最新推荐文章于 2024-02-02 10:51:06 发布

SinGaln

最新推荐文章于 2024-02-02 10:51:06 发布

阅读量392

点赞数 1

分类专栏：自然语言处理（NLP）深度学习（deep learning）文章标签： tensorflow MNIST

本文链接：https://blog.youkuaiyun.com/qq_27492735/article/details/89146226

版权

自然语言处理（NLP）同时被 2 个专栏收录

16 篇文章

订阅专栏

深度学习（deep learning）

13 篇文章

订阅专栏

使用TensorFlow进行简单MNIST数据集分类任务，未采用CNN或RNN，而是应用dropout，实验结果显示效果良好。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

import tensorflow as tf
import numpy as np
from tensorflow.examples.tutorials.mnist import input_data
import matplotlib.pyplot as plt

mnist = input_data.read_data_sets('MNIST_data/',one_hot=True)
batch_size = 50
learning_rate = 0.001
keep_prob = tf.placeholder(tf.float32)
n_batch = mnist.train.num_examples // batch_size
#import data

train_images = mnist.train.images
train_labels = mnist.train.labels
test_images = mnist.test.images
test_labels = mnist.test.labels
print("Loading data...")
print("training start....")
#construct graph
x_data = tf.placeholder(tf.float32,[None,784])
y_data = tf.placeholder(tf.float32,[None,10])


#initial variable
w1 =tf.Variable(tf.truncated_normal([784,600],stddev=0.1))
b1 = tf.Variable(tf.zeros([600])+0.1)
pre1 = tf.nn.tanh(tf.matmul(x_data,w1)+b1)
dro_pre1 = tf.nn.dropout(pre1,keep_prob)

w2 =tf.Variable(tf.truncated_normal([600,400],stddev=0.1))
b2 = tf.Variable(tf.zeros([400])+0.1)
pre2 = tf.nn.tanh(tf.matmul(pre1,w2)+b2)
dro_pre2 = tf.nn.dropout(pre2,keep_prob)

w3 =tf.Variable(tf.truncated_normal([400,200],stddev=0.1))
b3 = tf.Variable(tf.zeros([200])+0.1)
pre3 = tf.nn.tanh(tf.matmul(pre2,w3)+b3)
dro_pre3 = tf.nn.dropout(pre3,keep_prob)

w4 =tf.Variable(tf.truncated_normal([200,100],stddev=0.1))
b4 = tf.Variable(tf.zeros([100])+0.1)
pre4 = tf.nn.tanh(tf.matmul(pre3,w4)+b4)
dro_pre4 = tf.nn.dropout(pre4,keep_prob)

w5 =tf.Variable(tf.truncated_normal([100,10],stddev=0.1))
b5 = tf.Variable(tf.zeros([10])+0.1)
pre5 = tf.nn.softmax(tf.matmul(pre4,w5)+b5)
# dro_pre5 = tf.nn.dropout(pre5,keep_prob)
loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(labels=y_data,logits=pre5))

correct_pre = tf.equal(tf.argmax(y_data,1),tf.argmax(pre5,1))

accuracy = tf.reduce_mean(tf.cast(correct_pre,tf.float32))
#梯度下降
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(loss)

#innitial variable
init = tf.global_variables_initializer()
epoch_ = []
t_acc = []
with tf.Session() as sess:
    sess.run(init)
    for epoch in range (51):
            epoch_.append(epoch)
        # if epoch % 2 == 0:
            for batch in range(n_batch):
                batch_x,batch_y = mnist.train.next_batch(batch_size)
                sess.run(optimizer,feed_dict={x_data:batch_x,y_data:batch_y,keep_prob:0.7})

            train_acc = sess.run(accuracy,feed_dict={x_data:train_images,y_data:train_labels,keep_prob:0.7})
            test_acc = sess.run(accuracy,feed_dict={x_data:test_images,y_data:test_labels,keep_prob:0.7})
            t_acc.append(test_acc)
            print("Iter:" + str(epoch) + ", Training Accuracy:" + str(train_acc)+ ", Testing Accuracy:" + str(test_acc))

plt.figure()
plt.plot(epoch_,t_acc,'ro--',lw = 1)
plt.show()
plt.savefig('Test06 image')

因为是个练手程序，所以没有使用CNN或RNN，就是一个加了dropout的小程序，效果还行。