RFC: MultiEncoder add padding param separating various encoders' outputs

[breznak]

I'm currently experimenting with multiencoders, would it be useful to add a padding(=0) pattern which separates outputs of various sub-encoders in the final array?

Example: MultiEncoder.__init__(self, encoderDescriptions=None, padding=0)
Now: {EncoderOutput1 | EncoderOutput2 | ...}
After (padding=3): {EncoderOutput1 | 000 | EncoderOutput2 | 000 |...}

motivation is that the edges of neighboring encoders would make connected segment's synapses confused (the stable separation pattern should suppress that).

If you find it useful, can someone hack it up please? Im a bit lost in the MultiEnc code..

Thanks, breznak

[rhyolight]

Comments @scottpurdy or @chetan51?

[chetan51]

@breznak Are you seeing an issue with an experiment, or is this theoretical?

[scottpurdy]

This would only have an impact if you are using topology, which we generally don't do with MultiEncoder. I'm not sure there is much of a use case for it.