Reproduce “Near-Optimal Regret Bounds for UCB-Style Gaussian Process Bandits via Data Thinning” 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
Shion Takeno, Shogo Iwazaki, Ichiro Takeuchi
The authors introduce GP-DT-UCB and GP-DT-PE-UCB for optimization. The algorithms add Gaussian noise to observations, using one thinned data stream for adaptive UCB-style query selection and another for recommendation or elimination; the noise is conditionally Gaussian given inputs. Under finite-domain, bounded-RKHS-norm and Gaussian-noise assumptions, they prove near-optimal simple- and cumulative-regret bounds, matching lower bounds up to polylogarithmic factors.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.