Validation of the assumed 'SiMPle-likeness'

[mimbres]

Assumption:
If we replace RNNs with max-pooling, our network(NOT trainable) will be same as SiMPLe.

Let's examine this!