Reproduce “Beyond Monotonicity: Stochastic Online Optimization under Semi-Bandit Feedback” 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
Tsubasa Harada, Yasushi Kawase, Hanna Sumita
The authors present an online learning method that removes the need for stochastic-dominance monotonicity. Each decision uses an optimistic expectation chosen separately over confidence sets for the next variable, allowing optimism to depend on observed history. For rewards in [0,1], regret is \tilde O(\sqrt{ST}) when S is finite; continuous distributions are covered only when level-set complexity is finite.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.