Reproduce “Kernel $ε$-Greedy for Multi-Armed Bandits with Covariates” 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
Sakshi Arya, Bharath K. Sriperumbudur
The authors develop an epsilon-greedy treatment-allocation method with separate kernel models for each treatment and inverse-probability weighting to account for unequal assignment rates. The paper proves regret bounds under stated assumptions. For two arms in a finite-dimensional RKHS with a margin condition, expected cumulative regret is O(T^{1/2+ε}) for any ε>0; without the condition, the rate is O(T^{2/3}).
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.