Open GuillaumeLam opened 2 years ago
I never got around to a full implementation of STDP. That should be handled after the simulation completes, but I didn't build any sort of opinionated "how to update weights" abstraction in
I am currently looking into which STDP algorithm to follow. The updates could be after each simulation, but would it be more realistic if the updates were after each spike of a neuron? This would for sure incur a larger computational cost, however I am just scared that the per sim updates will be too spaced out for proper unsupervised learning...
Hello, I am very interested in implementing different STDP rules for DL tasks. Does anybody have a simple working example of an STDP rule that you want to share?
I am planning to add code for STDP in the next couple of days. Did you have an advice/experience in adding STDP capabilities to the framework?