Is there currently a way to write a learning that gets two different error signals as input?
You could pull this commit from the generic learning rule PR, which allowed learning rules to define arbitrary size_in
.
That doesn’t seem to work properly because the size_in
also defines one dimension of the the delta
signal (which is different).
In case anyone else is reading this and looking for that functionality: I put together an implementation in nengo/nengo#1307.