Closed studywolf closed 11 years ago
So Eric (Hunse) and I just talked with Terry, I think we have a good idea of how to go about a first stab at this, we'll put something together and update asap!
Alright, added network.learn (for api consistency), which calls ensemble.add_learned_termination, which adds pre-spikes * weight_matrix to the input_current calculation (by adding it through add_filtered_input) and then creates a learned_termination, which (depending on the learned_termination subclass implementation) will be in charge of modifying the weight matrix through some calculation involving the pre_spikes, post_spikes, and error ensemble/input/simplenode.
All of learning.