Spike timing and environment dependent optimisation
Ismaeel Ramzan
Abstract
We present a method for training spiking networks to spike at precise times called ReSuMe++. We compare this method with surrogate gradient descent using the Adam optimiser training on artificially generated data. We show that we are able to significantly improve upon the previous supervised STDP approach we extend but are not currently able to match the performance of the Adam optimiser doing surrogate gradient descent to solve our test problem.
Chat is not available.
Successful Page Load