Reproduce “A Non-Asymptotic Moreau Envelope Theory for High-Dimensional Generalized Linear Models” 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
Lijia Zhou, Frederic Koehler, Pragya Sur, Danica J. Sutherland, Nathan Srebro
The authors develop generalization bounds for interpolating predictors under potentially misspecified labels. They smooth the loss with a Moreau envelope, separating low-dimensional signal error from complexity in other directions. For square loss, the envelope is a rescaling, enabling bounds on population test loss that extend optimistic rates to nonlinear trends and heteroskedasticity. The theory assumes Gaussian features and concentration conditions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.