Reproduce “Training Deep Normalization-Free Spiking Neural Networks with Lateral Inhibition” 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
Peiyu Liu, Jianhao Ding, Zhaofei Yu
The authors introduce DeepEISNN, a deep spiking network that replaces explicit normalization layers with interacting excitatory and inhibitory populations. E-I Init uses first-batch statistics once, while E-I Prop stabilizes divisive inhibition and scales gradients; training still uses backpropagation. On CIFAR-10, VGG-19 reaches 90.93±0.33% top-1 accuracy, versus 88.28±0.27% for DANN, an ANN, in a nonmatched comparison that does not isolate the cause.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.