Reproduce “Optimistic Rates: A Unifying Theory for Interpolation Learning and Regularization in Linear Regression” 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, Danica J. Sutherland, Nathan Srebro
The paper bounds population squared loss in well-specified Gaussian linear regression using both observed training loss and predictor-class complexity. Its optimistic guarantee adapts complexity to predictor scale and, up to lower-order terms, has leading factor 1. It covers imperfect fits as well as interpolators; under stated conditions, OLS expected loss is σ²(n−1)/(n−d−1) for d<n−1, not a guarantee for arbitrary distributions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.