Keras 2: Using lambda function in "Merge" layers

Question

I am trying to implement this merge layer:

policy = merge([out1, out2], mode = lambda x: x[0]-K.mean(x[0])+x[1], output_shape = (out_node,))

However, "merge" is no longer present in Keras 2. You can only access public standarized "Merge" layers, such as Add, Multiply, Dot.

How can I implement this function in Keras 2? I thought about making two merge layers but I have no idea how to implement that, especially because of the "K.mean" part.

For reference, here are the imports:

from keras.layers import merge
from keras import backend as K

So you want to subtract the mean of `out1` from it and then add the result to `out2`, element-wise? — today, Apr 20 '19 at 15:59

score 7 · Accepted Answer · answered Apr 20 '19 at 17:14

7

You can simply do this using a Lambda layer:

from keras import backend as K
from keras.layers import Lambda

policy = Lambda(lambda x: x[0] - K.mean(x[0]) + x[1])([out1, out2])

answered Apr 20 '19 at 17:14

today

32,602
8
95
115

Keras 2: Using lambda function in "Merge" layers

1 Answers1