Reproduce “Gradient Descent for Spiking Neural Networks” 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

Dongsung Huh, Terrence J. Sejnowski

The authors make supervised learning in recurrent spiking networks differentiable by smoothing synaptic current within a narrow voltage “active zone,” preserving charge across a full threshold crossing, and deriving exact gradients for the resulting model. They train networks directly with backpropagation through time. Demonstrations include delayed-memory XOR and auto-encoding; selected trials show task responses, not aggregate accuracy or broad generalization.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.