tensorflow 1.x practical tutorial (6) - simple classification model

Target

This article aims to introduce the introductory knowledge points and practical examples of tensorflow. I hope that all novice students can be proficient in tensorflow related operations after learning.

Simple classification model code

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

mnist = input_data.read_data_sets("MNIST", one_hot=True) # 读数据
batch_size = 64
n_batchs = mnist.train.num_examples // batch_size

x = tf.placeholder(dtype=tf.float32, shape=[None, 784], name='x') # 因为每张图片数据是 28*28=784 维的
y = tf.placeholder(dtype=tf.float32, shape=[None, 10], name='y')  # 因为一共有 10 种类别的图片

# softmax((x * w) + b)
w = tf.Variable(tf.ones(shape=[784, 10]))
b = tf.Variable(tf.zeros(shape=[10]))
predict = tf.nn.softmax(tf.matmul(x, w) + b)

loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=predict, labels=y))
opt = tf.train.GradientDescentOptimizer(0.1).minimize(loss)
init = tf.global_variables_initializer()

correct = tf.equal(tf.argmax(predict,1), tf.argmax(y,1))
accuracy = tf.reduce_mean(tf.cast(correct, tf.float32)) # 计算准确率

with tf.Session() as sess:
    sess.run(init)
    total_batch = 0
    last_batch = 0
    best = 0
    for epoch in range(100):
        for _ in range(n_batchs):
            xx,yy = mnist.train.next_batch(batch_size)
            sess.run(opt, feed_dict={x:xx, y:yy})
        loss_value, acc = sess.run([loss, accuracy], feed_dict={x:mnist.test.images, y:mnist.test.labels})
        # 始终打印最好的准确率信息
        if acc > best:
            best = acc
            last_batch  = total_batch
            print('epoch:%d, loss:%f, acc:%f' % (epoch, loss_value, acc))
        if total_batch - last_batch > 5: # 训练早停条件
            print('when epoch-%d early stop train'%epoch)
            break
        total_batch += 1
        
复制代码

output result

Extracting MNIST/train-images-idx3-ubyte.gz
Extracting MNIST/train-labels-idx1-ubyte.gz
Extracting MNIST/t10k-images-idx3-ubyte.gz
Extracting MNIST/t10k-labels-idx1-ubyte.gz
WARNING:tensorflow:From <ipython-input-18-ec0f1616d772>:16: softmax_cross_entropy_with_logits (from tensorflow.python.ops.nn_ops) is deprecated and will be removed in a future version.
Instructions for updating:

Future major versions of TensorFlow will allow gradients to flow
into the labels input on backprop by default.

See tf.nn.softmax_cross_entropy_with_logits_v2.

epoch:0, loss:1.697559, acc:0.819900
epoch:1, loss:1.627650, acc:0.887500
epoch:2, loss:1.604011, acc:0.897200
epoch:3, loss:1.592221, acc:0.902300
epoch:4, loss:1.585058, acc:0.904700
epoch:5, loss:1.579867, acc:0.907700
epoch:6, loss:1.575740, acc:0.909700
epoch:7, loss:1.572829, acc:0.911200
epoch:8, loss:1.570307, acc:0.912600
epoch:9, loss:1.567902, acc:0.913100
epoch:10, loss:1.565990, acc:0.913900
epoch:11, loss:1.564570, acc:0.916100
epoch:13, loss:1.561729, acc:0.917800
epoch:14, loss:1.560736, acc:0.917900
epoch:15, loss:1.559514, acc:0.918600
epoch:17, loss:1.557875, acc:0.919600
epoch:18, loss:1.557073, acc:0.920100
epoch:21, loss:1.554998, acc:0.920500
epoch:22, loss:1.554592, acc:0.920700
epoch:23, loss:1.553998, acc:0.921500
epoch:24, loss:1.553378, acc:0.922100
epoch:28, loss:1.551517, acc:0.922400
epoch:29, loss:1.551527, acc:0.922700
epoch:31, loss:1.550692, acc:0.923000
epoch:32, loss:1.550284, acc:0.923200
epoch:33, loss:1.550164, acc:0.923300
epoch:34, loss:1.549571, acc:0.923600
epoch:35, loss:1.549563, acc:0.923700
epoch:38, loss:1.548744, acc:0.923800
epoch:39, loss:1.548406, acc:0.924700
epoch:41, loss:1.547895, acc:0.924800
epoch:45, loss:1.547032, acc:0.925300
epoch:49, loss:1.546252, acc:0.925900
epoch:51, loss:1.545930, acc:0.926400
epoch:56, loss:1.545088, acc:0.926700
epoch:59, loss:1.544781, acc:0.927400
epoch:65, loss:1.544077, acc:0.927500
epoch:66, loss:1.543733, acc:0.927800
epoch:70, loss:1.543496, acc:0.928100
epoch:76, loss:1.542884, acc:0.928300
epoch:80, loss:1.542315, acc:0.928600
when epoch-86 early stop train    
复制代码

Reference in this article

Reference for this article: blog.csdn.net/qq_19672707…