Reproduce “EigenGame: PCA as a Nash Equilibrium” 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

Ian Gemp, Brian McWilliams, Claire Vernade, Thore Graepel

EigenGame recasts PCA as a game in which each vector earns variance reward and later vectors are penalized for overlap with earlier ones. Its decentralized hierarchy updates components separately and passes messages downstream. It recovered 32 ResNet-200 activation components in about 20 million dimensions using 32 TPUv3s in nine hours, demonstrating scale rather than a general speed advantage.

Summary by alphaXiv.

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