“Training spiking neural networks using lessons from deep learning using shiny spiking learning rules”.
During this CONECT seminar, Jason Eshraghian will present his recent work on “Making spiking neural networks do useful things”
This presentation will dive into how spiking neural networks can be trained to accomplish practical engineering problems. We will provide an overview of the various learning rules that have emerged over the past several decades, along with a few large-scale applications we’ve achieved with spike-based computation. This involves our spike-based language model, SpikeGPT, and our open-source Python library that adopts gradient-based optimization into spike-based models, snnTorch.
In addition, we had a master class in the morning on snnTorch, get the notebook (upon request).