zoukankan      html  css  js  c++  java
  • 基于tensorflow的CNN卷积神经网络对手写字MNIST数据集的分类器

    这篇用一种更简洁的方式写CNN神经网络

    上代码:

    import tensorflow as tf
    from tensorflow.examples.tutorials.mnist import  input_data
    mnist = input_data.read_data_sets('MNIST_data', one_hot=True)
    #define conv layer
    def conv_layer(inputs, channels_in, channels_out):
        w = tf.Variable(tf.truncated_normal([5,5,channels_in, channels_out], stddev=0.1))
        b = tf.Variable(tf.constant(0.1, shape=[channels_out]))
        conv = tf.nn.conv2d(inputs, w, strides=[1, 1, 1, 1], padding='SAME')  #调用conv2d
        act = tf.nn.relu(conv + b)    # conv + bias
        return act
    #define fully connection layer
    def fc_layer(inputs, channels_in, channels_out, activate_function=None):
        w = tf.Variable(tf.truncated_normal([channels_in, channels_out]))
        b = tf.Variable(tf.zeros([channels_out])+0.1)
        fc = tf.matmul(inputs, w) + b
        if activate_function is None:
            res = fc
        else:
            res = activate_function(fc)
        return res
    x = tf.placeholder(tf.float32, [None, 784])
    y = tf.placeholder(tf.float32, [None, 10])
    x_image = tf.reshape(x, [-1, 28, 28, 1])    
    keep_prob = tf.placeholder(tf.float32)
    conv1 = conv_layer(x_image, 1, 32)
    pool1 = tf.nn.max_pool(conv1, ksize=[1, 1, 1, 1], strides=[1, 2, 2, 1], padding='SAME')    # 14*14*32
    
    conv2 = conv_layer(pool1, 32, 64)
    pool2 = tf.nn.max_pool(conv2, ksize=[1, 1, 1, 1], strides=[1, 2, 2, 1], padding='SAME')    
    flattended = tf.reshape(pool2, [-1, 7*7*64])
    fc1 = fc_layer(flattended, 7 * 7 * 64, 1024, tf.nn.relu)
    fc1 = tf.nn.dropout(fc1, keep_prob)
    logits =  fc_layer(fc1, 1024, 10)
    cross_entropy = tf.reduce_mean(
                            tf.nn.softmax_cross_entropy_with_logits(logits=logits, labels=y)
    )
    train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
    correct_prediction = tf.equal(tf.argmax(logits, 1), tf.argmax(y, 1))
    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
    init = tf.global_variables_initializer()
    sess = tf.Session()
    sess.run(init)
    for i in range(10000):
        #batch = mnist.train.next_batch(100)
        images, labels = mnist.train.next_batch(100)
        sess.run(train_step, feed_dict={x:images, y:labels,keep_prob:1})
        if i%50==0:
            print(sess.run(accuracy, feed_dict={x:mnist.test.images[:1000], y:mnist.test.labels[:1000], keep_prob:1}))
            #print(sess.run(accuracy, feed_dict={x:batch[0], y:batch[1]}))
    sess.close()

    1.这段代码封装了CNN的每一层

    2.这段代码还用到了droup out 防止过拟合

    3.不知道为什么,这样的写法似乎比不封装要慢,最终准确率也无法到达98 99 

  • 相关阅读:
    ASP扫盲学习班第六课 添加新保存的数据
    三级联动菜单免刷新加载
    ASP按定制格式导出word文件的完美解决思路
    通用客户资源管理系统(客户资料保护库)
    SmR 通用信息采集系统(新闻小偷)
    ASP扫盲学习班第三课 程序界面的设计
    我的新作品(单身贵族网全站)
    《将博客搬至CSDN》
    我的新作品
    asp函数大全
  • 原文地址:https://www.cnblogs.com/francischeng/p/9894588.html
Copyright © 2011-2022 走看看