内容来自mooc人工智能实践第五讲
一、MNIST数据集一些用到的基础函数语法
############ warm up ! ############
# 导入imput_data模块
from tensorflow.examples.tutorials.mnist import input_data
# 加载数据集,以读热码的形式存取
mnist = input_data.read_data_sets('./data/', one_hot = True)
# 打印训练集、验证集、测试集所含有的样本数
print("train data size:", mnist.train.num_examples)
print("validation data size:", mnist.validation.num_examples)
print("test data size:", mnist.test.num_examples)
# 查看训练集中指定编号的标签或图片数据
mnist.train.labels[0]
mnist.train.images[0]
# 将训练集中一定batchsize的数据和标签赋给左边的变量
xs, ys = minist.train.next_batch(BATCH_SIZE)
# 打印形状
print("xs shape: ", xs.shape)
print("ys shape: ", ys.shape)
# 从集合中取全部变量,生成一个列表
tf.get_collection("")
# 列表内对应元素相加
tf.add_n([])
# 将x转化为指定类型
tf.cast(x, dtype)
# 对比两个矩阵或者向量的每个元素,对应元素相等时依次返回True,否则False
A = [[1,3,4,5,6]]
B = [[1,3,4,3,2]]
with tf.Session() as sess:
print(sess.run(tf.equal(A, B)))
# 求均值
# 若不指定第二个参数,则在所有元素中求平均值
# 若第二个参数0,则在第一维元素上取平均值
# 若第二个参数1,则在第二维元素上求平均值
tf.reduce_mean(x, axis)
# 返回axis指定的维度中,列表x最大值对应的索引号
tf.argmax(x, axis)
# 拼接路径
import os
os.path.join("home", "name") # 返回home/name
# 按指定拆分符,对字符串切片,返回分割后的列表(字符串)
# 用于从一个文件中读取global step的值
'./mode/mnist_model-1001'.split('/')[-1].split('-')[-1] # 返回1001
# 用于复现已经定义好了的神经网络
with tf.Gragh().as_default() as g: # 其内定义的节点在计算图g中
###### 模型的保存 ######
# 反向传播中,一般每隔一定轮数把神经网络模型保存下来
# 保存三个文件
# 1.当前图结构的.meta文件
# 2.当前参数名的.index文件
# 3.当前参数的.data文件
saver = tf.train.Saver() # 实例化saver对象
with tf.Session() as sess: # 在with结构中循环一定轮数时,保存模型到当前会话
for i in range(STEPS):
if i % 轮数 == 0: # 拼接成./MODEL_SAVE_PATH/MODEL_NAME-global_step路径
saver.save(sess, os.path.join(MODEL_SAVE_PATH, MODEL_NAME), global_step = global_step)
# 将神经网络模型中的所有参数等信息保存到指定路径中,并在存放网络模型的文件夹名称中注明保存模型时的训练轮数
# 测试网络效果时,需要将训练好的神经网络模型加载
with tf.Session() as sess:
ckpt = tf.train.get_checkpoint_state(存储路径)
if ckpt and ckpt.model_checkpoint_path: #若ckpt和保存的模型在指定路径中存在
saver.restore(sess, ckpt.model_checkpoint_path) #则将保存的神经网络模型加载到当前会话中
# 加载模型中参数的滑动平均值
# 保存模型时,若模型中采用了滑动平均,则参数的滑动平均值会保存在相应文件中
ema = tf.train.ExponentialMovingAverage(滑动平均基数)
ema_restore = ema.variables_to_restore()
# 实例化可以还原滑动平均值的saver对象
saver = tf.train.Saver(ema_restore)
# 神经网络模型准确率的评估方法
# y 表示在一组batch_size大小的数据上,神经网络模型的预测结果
# y.shape = [batch_size, 10]
# 判断预测记过张量和实际标签张量的每个维度是否相等
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
# 将布尔值转化为实数型
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
二、测试过程test.py及主函数
######## test.py ##########
def test(mnist):
with tf.Gragh().as_default() as g:
#占位
x = tf.placeholder(dtype, shape)
y_= tf.placeholder(dtype, shape)
# 前向传播,预测结果y
y = mnist_forward.forward(x, None)
# 实例化可以还原滑动平均的saver
ema = tf.train.ExponentialMovingAverage(滑动衰减率)
ema_restore = ema.variables_to_restore()
# 实例化可以还原滑动平均值的saver对象
saver = tf.train.Saver(ema_restore)
# 计算正确率
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
while True:
with tf.Session() as sess:
# 加载训练好的模型
ckpt = tf.train.get_checkpoint_state(存储路径)
#若ckpt和保存的模型在指定路径中存在
if ckpt and ckpt.model_checkpoint_path:
# 恢复会话
saver.restore(sess, ckpt.model_checkpoint_path)
# 恢复轮数
global_ste = ckpt.model_checkpoint_path.split('/')[-1].split('-')[-1]
# 计算准确率
accuracy_score = sess.run(accuracy, feed_dict = {x:测试数据, y_:测试数据标签})
# 打印提示
print("after %s training steps, test_accuracy = %g" %(global_step, accuracy_score))
#如果没有模型
else:
print("no checkpoint file found")
return
######## main function #############
def main():
mnist = input_data.read_data_sets('./data/', one_hot = True)
# 调用定义好的测试函数
test(mnist)
if __name__ == '__main__':
main()
三、完整代码
- ①
mnist_forward.py
# mnist_forward.py
# coding: utf-8
import tensorflow as tf
INPUT_NODE = 784
OUTPUT_NODE = 10
LAYER1_NODE = 500
# 给w赋初值,并把w的正则化损失加到总损失中
def get_weight(shape, regularizer):
w = tf.Variable(tf.truncated_normal(shape, stddev = 0.1))
if regularizer != None: tf.add_to_collection('losses', tf.contrib.layers.l2_regularizer(regularizer)(w))
return w
# 给b赋初值
def get_bias(shape):
b = tf.Variable(tf.zeros(shape))
return b
def forward(x, regularizer):
w1 = get_weight([INPUT_NODE, LAYER1_NODE], regularizer)
b1 = get_bias([LAYER1_NODE])
y1 = tf.nn.relu(tf.matmul(x, w1) + b1)
w2 = get_weight([LAYER1_NODE, OUTPUT_NODE], regularizer)
b2 = get_bias([OUTPUT_NODE])
y = tf.matmul(y1, w2) + b2 #输出层不通过激活函数
return y
- ②
mnist_backward.py
# mnist_backward.py
# coding: utf-8
import tensorflow as tf
# 导入imput_data模块
from tensorflow.examples.tutorials.mnist import input_data
import mnist_forward
import os
# 定义超参数
BATCH_SIZE = 200
LEARNING_RATE_BASE = 0.1 #初始学习率
LEARNING_RATE_DECAY = 0.99 # 学习率衰减率
REGULARIZER = 0.0001 # 正则化参数
STEPS = 50000 #训练轮数
MOVING_AVERAGE_DECAY = 0.99
MODEL_SAVE_PATH = "./model/"
MODEL_NAME = "mnist_model"
def backward(mnist):
# placeholder占位
x = tf.placeholder(tf.float32, shape = (None, mnist_forward.INPUT_NODE))
y_ = tf.placeholder(tf.float32, shape = (None, mnist_forward.OUTPUT_NODE))
# 前向传播推测输出y
y = mnist_forward.forward(x, REGULARIZER)
# 定义global_step轮数计数器,定义为不可训练
global_step = tf.Variable(0, trainable = False)
# 包含正则化的损失函数
# 交叉熵
ce = tf.nn.sparse_softmax_cross_entropy_with_logits(logits = y, labels = tf.argmax(y_, 1))
cem = tf.reduce_mean(ce)
# 使用正则化时的损失函数
loss = cem + tf.add_n(tf.get_collection('losses'))
# 定义指数衰减学习率
learning_rate = tf.train.exponential_decay(
LEARNING_RATE_BASE,
global_step,
mnist.train.num_examples/BATCH_SIZE,
LEARNING_RATE_DECAY,
staircase = True)
# 定义反向传播方法:包含正则化
train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss, global_step = global_step)
# 定义滑动平均时,加上:
ema = tf.train.ExponentialMovingAverage(MOVING_AVERAGE_DECAY, global_step)
ema_op = ema.apply(tf.trainable_variables())
with tf.control_dependencies([train_step, ema_op]):
train_op = tf.no_op(name = 'train')
# 实例化saver
saver = tf.train.Saver()
# 训练过程
with tf.Session() as sess:
# 初始化所有参数
init_op = tf.global_variables_initializer()
sess.run(init_op)
# 循环迭代
for i in range(STEPS):
# 将训练集中一定batchsize的数据和标签赋给左边的变量
xs, ys = mnist.train.next_batch(BATCH_SIZE)
# 喂入神经网络,执行训练过程train_step
_, loss_value, step = sess.run([train_op, loss, global_step], feed_dict = {x: xs, y_: ys})
if i % 1000 == 0: # 拼接成./MODEL_SAVE_PATH/MODEL_NAME-global_step路径
# 打印提示
print("after %d steps, loss on traing batch is %g" %(step, loss_value))
saver.save(sess, os.path.join(MODEL_SAVE_PATH, MODEL_NAME), global_step = global_step)
def main():
mnist = input_data.read_data_sets('./data/', one_hot = True)
# 调用定义好的测试函数
backward(mnist)
# 判断python运行文件是否为主文件,如果是,则执行
if __name__ == '__main__':
main()
- ③
mnist_test.py
# coding:utf-8
# mnist_test.py
# 延时
import time
import tensorflow as tf
# 导入imput_data模块
from tensorflow.examples.tutorials.mnist import input_data
import mnist_forward
import mnist_backward
import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2' #hide warnings
# 程序循环间隔时间5秒
TEST_INTERVAL_SECS = 5
def test(mnist):
# 用于复现已经定义好了的神经网络
with tf.Graph().as_default() as g: # 其内定义的节点在计算图g中
# placeholder占位
x = tf.placeholder(tf.float32, shape=(None, mnist_forward.INPUT_NODE))
y_ = tf.placeholder(tf.float32, shape=(None, mnist_forward.OUTPUT_NODE))
# 前向传播推测输出y
y = mnist_forward.forward(x, None)
# 实例化带滑动平均的saver对象
# 这样,所有参数在会话中被加载时,会被复制为各自的滑动平均值
ema = tf.train.ExponentialMovingAverage(mnist_backward.MOVING_AVERAGE_DECAY)
ema_restore = ema.variables_to_restore()
saver = tf.train.Saver(ema_restore)
# 计算正确率
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
while True:
with tf.Session() as sess:
# 加载训练好的模型,也即把滑动平均值赋给各个参数
ckpt = tf.train.get_checkpoint_state(mnist_backward.MODEL_SAVE_PATH)
#若ckpt和保存的模型在指定路径中存在
if ckpt and ckpt.model_checkpoint_path:
# 恢复会话
saver.restore(sess, ckpt.model_checkpoint_path)
# 恢复轮数
global_step = ckpt.model_checkpoint_path.split('/')[-1].split('-')[-1]
# 计算准确率
accuracy_score = sess.run(accuracy, feed_dict={x: mnist.test.images, y_: mnist.test.labels})
# 打印提示
print("after %s training steps, test accuracy = %g" % (global_step, accuracy_score))
#如果没有模型
else:
print("no checkpoint file found")
return
time.sleep(TEST_INTERVAL_SECS)
def main():
mnist = input_data.read_data_sets('./data/', one_hot=True)
# 调用定义好的测试函数
test(mnist)
if __name__ == '__main__':
main()
从终端运行结果可以看出,随着训练轮数的增加,网络模型的损失函数值不断降低,并且在测试集上的准确率在不断提升,有较好的泛化能力。
从上图结果可以看出,最终迭代后准确率基本稳定不变了。