文章目录
- 一、数据集mnist下载
- 二、0-9分类训练
说明:
- 本文主要是将TensorFlow例程拿过来解释了一下
- 首先是数据的下载,注意下载的时候可能会出错,但是网上有很多其他的下载网址可用
- 1 加载mnist数据集
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True) - 2 清除图形堆栈
tf.reset_default_graph() - 3 参数和输入输出占位
tf.Variable和tf.placeholder - 4 配置前向输出
tf.nn.softmax(tf.matmul(x, W) + b) - 5 配置交叉熵(或loss)与梯度下降
tf.reduce_mean, tf.train.GradientDescentOptimizer(learning_rate).minimize(cost) - 6 设置训练的参数设置
- 7 启动训练
- 8 测试
- 9 保存训练模型
- 10 加载训练模型
- 11 重新测试
一、数据集mnist下载
官网有下载但是有出错的情况, 这里给出直接下载
import tensorflow as tf #导入tensorflow库
from tensorflow.examples.tutorials.mnist import input_data
import pylab
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True) # 下载并加载数据
# print('训练集:', mnist.train.images) # 训练集
print('训练集shape:', mnist.train.images.shape) # 输入数据打印shape: (55000, 784)
im = mnist.train.images[1]
im = im.reshape(-1, 28)
pylab.imshow(im)
pylab.show()
print('测试集shape:', mnist.test.images.shape) # 输入数据打印shape: (10000, 784) 测试集
print('输入数据打印shape:', mnist.validation.images.shape)
print("训练数据总数:" , mnist.train.num_examples) # 550 00
print("测试数据总数:" , mnist.test.num_examples) # 1w
Extracting MNIST_data/train-images-idx3-ubyte.gz
Extracting MNIST_data/train-labels-idx1-ubyte.gz
Extracting MNIST_data/t10k-images-idx3-ubyte.gz
Extracting MNIST_data/t10k-labels-idx1-ubyte.gz
训练集shape: (55000, 784)
测试集shape: (10000, 784)
输入数据打印shape: (5000, 784)
训练数据总数: 55000
测试数据总数: 10000
二、0-9分类训练
import tensorflow as tf #导入tensorflow库
from tensorflow.examples.tutorials.mnist import input_data
import pylab
# 1 加载数据
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
print("训练数据总数:" , mnist.train.num_examples)
print(mnist.test.nu)
# 2 函数用于清除默认图形堆栈并重置全局默认图形
tf.reset_default_graph()
# 3 tf Graph Input 输入输出占位
x = tf.placeholder(tf.float32, [None, 784]) # mnist data维度 28*28=784
y = tf.placeholder(tf.float32, [None, 10]) # 0-9 数字=> 10 classes
# 4 Set model weights 权重占位,
W = tf.Variable(tf.random_normal([784, 10]))
b = tf.Variable(tf.zeros([10]))
# 5 前向输出
pred = tf.nn.softmax(tf.matmul(x, W) + b) # Softmax分类 # 激活函数,输出为总输出的概率, 总和为1,
# 6 生成的pred与样本标签y进行一次交叉熵运算, 然后在取平均值, 注意这里的交叉熵运行在上一节已经提到过
# 将这个结果作为一次正向传播的误差, 通过梯度下降的优化方法找到能够使这个误差最小化的b,w偏移量
# 更细b,w参数, 使其调整为合适的参数
cost = tf.reduce_mean(-tf.reduce_sum(y*tf.log(pred), reduction_indices=1))
# 7 使用梯度下降优化器
learning_rate = 0.01 # 学习率, 对于梯度下降法的输入参数
optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost)
# 8 训练参数设置
training_epochs = 25
batch_size = 100
display_step = 1
saver = tf.train.Saver()
model_path = "log/521model.ckpt"
# 9 启动session
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())# Initializing OP
# 启动循环开始训练
for epoch in range(training_epochs): # 循环25次 25*550次循环,每次循环喂100
avg_cost = 0.
total_batch = int(mnist.train.num_examples/batch_size) # total_batch=55000/100=550 一次训练
# 注意下面的for是训练一遍数据的
# 遍历全部数据集
for _1 in range(total_batch): # 1 - 550
batch_xs, batch_ys = mnist.train.next_batch(batch_size) # 拿出来100个喂给训练兽
# Run optimization op (backprop) and cost op (to get loss value)
_2, c = sess.run([optimizer, cost], feed_dict={x: batch_xs,
y: batch_ys})
# Compute average loss 计算平均loss值,
avg_cost += (c / total_batch)
# 显示训练中的详细信息
if (epoch+1) % display_step == 0:
print ("Epoch:", '%04d' % (epoch+1), "cost=", "{:.9f}".format(avg_cost))
print( " Finished!")
# 测试 model
correct_prediction = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))
# 计算准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
print ("Accuracy:", accuracy.eval（{x: mnist.test.images, y: mnist.test.labels}))
# Save model weights to disk
save_path = saver.save(sess, model_path)
print("Model saved in file: %s" % save_path)
Epoch: 0001 cost= 8.740313732
...
Epoch: 0025 cost= 0.849729373
Finished!
Accuracy: 0.8294
Model saved in file: log/521model.ckpt
'''
#读取模型
print("Starting 2nd session...")
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
saver.restore(sess, model_path)
# 测试 model
correct_prediction = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))
# 计算准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
print ("Accuracy:", accuracy.eval（{x: mnist.test.images, y: mnist.test.labels}))
output = tf.argmax(pred, 1)
batch_xs, batch_ys = mnist.train.next_batch(2)
outputval,predv = sess.run([output,pred], feed_dict={x: batch_xs})
print(outputval,predv,batch_ys)
im = batch_xs[0]
im = im.reshape(-1,28)
pylab.imshow(im)
pylab.show()
im = batch_xs[1]
im = im.reshape(-1,28)
pylab.imshow(im)
pylab.show()
'''
