Reproduce “Gaussian Partitioning for Reliable Learning of Multi-Index 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
Ilias Diakonikolas, Giannis Iakovidis, Daniel M. Kane, Mingchen Ma
The authors develop a partitioning method that lets Gaussian tools handle data altered by reweighting or conditioning. It divides data into thin regions and matches Gaussian moments along unrestricted directions, while smoothness controls discarded mass. For intersections of k halfspaces, positive-example learning has dimension dependence d^{O(log(k/ε))} for fixed parameters; its matching lower bound is only in the Statistical Query model.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.