Reproduce “Stochastic Local Winner-Takes-All Networks Enable Profound Adversarial Robustness” 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

Konstantinos P. Panousis, Sotirios Chatzis, Sergios Theodoridis

The authors test whether stochastic local winner-takes-all (LWTA) activations improve image classification under adversarial perturbations. Each small group samples one competing unit to pass its response onward, changing the network’s active route; the model is trained with PGD-based adversarial training. On CIFAR-10, a WideResNet-34-10 reaches 82.60% robust accuracy under AutoAttack, exceeding listed baselines, but this leaderboard comparison is not controlled.

Summary by alphaXiv.

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