How to implement a middle pool in Keras

Question

There seems to be no built-in mid-pool level support for RNN in Keras. Does anyone know how to wrap it?

http://deeplearning.net/tutorial/lstm.html

Answer 1

Keras has an AveragePooling1D layer for this. If you are using a graphical API, you should do something like:

 model.add_node(AveragePooling1D(...), inputs=['h0', 'h1', ..., 'hn'], merge_mode='concat', ...)