Reproduce “HoSNN: Adversarially-Robust Homeostatic Spiking Neural Networks with Adaptive Firing Thresholds” 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
Hejia Geng, Peng Li
HoSNN improves spiking neural network robustness by adjusting each neuron's firing threshold online toward a clean-training-data average membrane-potential trace. This reference is a dataset average, not activity for a particular test image. On CIFAR-10, with matched FGSM training, HoSNN achieves 43.33% accuracy under white-box PGD7 at ε = 8/255, versus 12.42% for LIF; the finding applies to tested settings, not arbitrary attacks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.