tensorflow中的EMA(指数平均数指标)

tensorflow 中的 ExponentialMovingAverage

看官方文档中的公式: 

shadowVariable=decay*shadowVariable+(1-decay)*variable

,就知道各代表什么意思了。

import tensorflow as tf
w = tf.Variable(1.0)
ema = tf.train.ExponentialMovingAverage(0.9)
update = tf.assign_add(w, 1.0)

with tf.control_dependencies([update]):
    #返回一个op,这个op用来更新moving_average
    ema_op = ema.apply([w])#这句和下面那句不能调换顺序
# 此op用来返回当前的moving_average
ema_val = ema.average(w)#参数不能是list，有点蛋疼

with tf.Session() as sess:
    tf.global_variables_initializer().run()
    for i in range(3):
        sess.run(ema_op)
        print(sess.run(ema_val))
# 创建一个时间序列 1 2 3 4
#输出：
#1.1      =0.9*1 + 0.1*2
#1.29     =0.9*1.1+0.1*3
#1.561    =0.9*1.29+0.1*4
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21

你可能会奇怪，明明 只执行三次循环， 为什么产生了 4 个数？ 
这是因为，当程序执行到 ema_op = ema.apply([w]) 的时候，如果 w 是 Variable， 那么将会用 w 的初始值初始化 ema 中关于 w 的 ema_value，所以 emaVal0=1.0。如果 w 是 Tensor的话，将会用 0.0 初始化。

官网中的示例：

# Create variables.
var0 = tf.Variable(...)
var1 = tf.Variable(...)
# ... use the variables to build a training model...
...
# Create an op that applies the optimizer.  This is what we usually
# would use as a training op.
opt_op = opt.minimize(my_loss, [var0, var1])

# Create an ExponentialMovingAverage object
ema = tf.train.ExponentialMovingAverage(decay=0.9999)

# Create the shadow variables, and add ops to maintain moving averages
# of var0 and var1.
maintain_averages_op = ema.apply([var0, var1])

# Create an op that will update the moving averages after each training
# step.  This is what we will use in place of the usual training op.
with tf.control_dependencies([opt_op]):
    training_op = tf.group(maintain_averages_op)
    # run这个op获取当前时刻 ema_value
    get_var0_average_op = ema.average(var0)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22

shadow variables are created with trainable=False。

使用 ExponentialMovingAveraged parameters

假设我们使用了ExponentialMovingAverage方法训练了神经网络， 在test阶段，如何使用 ExponentialMovingAveraged parameters呢？ 官网也给出了答案 
方法一：

# Create a Saver that loads variables from their saved shadow values.
shadow_var0_name = ema.average_name(var0)
shadow_var1_name = ema.average_name(var1)
saver = tf.train.Saver({shadow_var0_name: var0, shadow_var1_name: var1})
saver.restore(...checkpoint filename...)
# var0 and var1 now hold the moving average values
1
2
3
4
5
6
1
2
3
4
5
6

方法二：

#Returns a map of names to Variables to restore.
variables_to_restore = ema.variables_to_restore()
saver = tf.train.Saver(variables_to_restore)
...
saver.restore(...checkpoint filename...)
1
2
3
4
5
1
2
3
4
5

这里要注意的一个问题是，用于保存的saver可不能这么写，参考 http://blog.csdn.net/u012436149/article/details/56665612

参考资料

https://www./versions/master/api_docs/python/train/moving_averages