Reproduce “Robust and accelerated single-spike spiking neural network training with applicability to challenging temporal tasks” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Luke Taylor, Andrew King, Nicol Harper
FastSNN trains single-spike spiking neural networks by computing membrane activity across time in parallel, retaining only the first spike and learning membrane time constants. On spoken-digit classification, its model reaches 70.32% accuracy, close to 70.81% for a matched-size feedforward multi-spike model and above 44.50% with fixed time constants. FastSNN takes 11.27 seconds per epoch, versus 50.12 seconds for a sequential single-spike control.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.