Reproduce “A One-Sided McDiarmid-Type Inequality for Heavy Tails via Reverse KL Tensorization” 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
Hanuka Eshima, Amedeo Roberto Esposito, Andrea Paudice
The authors extend McDiarmid’s inequality to nonlinear functions of independent inputs by bounding conditional p-th moments of one-sided gaps between function values and coordinatewise ceilings. Reverse-KL tensorization yields an exponential upper-tail bound for 1<p≤2, without requiring two-sided bounded changes. They bound population risk by empirical risk and complexity terms under finite-moment and finite-complexity assumptions; this is not two-sided concentration.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.