Reproduce “Efficient Proper Agnostic Learning of Halfspaces” 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
Anindya De, Elchanan Mossel, Shivam Nadimpalli, Joe Neeman
The authors present a proper learner for agnostically learning affine halfspaces on the Boolean cube, using random restrictions and gradient averaging to find a low-dimensional feature space for its search. With membership queries, it achieves error at most OPT plus ε using O(n ε^-28) queries. Its runtime has an exponent of n independent of ε, but includes an exponential-in-1/ε term.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.