Reproduce “Un-regularizing: approximate proximal point and faster stochastic algorithms for empirical risk minimization” 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

Roy Frostig, Rong Ge, Sham M. Kakade, Aaron Sidford

APPA minimizes ERM by approximately solving a strongly regularized objective centered at the current estimate, then moving the center. Under the paper’s assumptions, its running-time bound scales with √κ rather than κ for SVRG. For squared-loss MNIST and CIFAR, APPA uses a tenfold larger penalty than SDCA while reaching a lower final ERM value; this finding is limited to those tasks.

Summary by alphaXiv.

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