A new research paper by Masahiro Kato introduces a novel strategy for best-arm identification in fixed-budget scenarios. The proposed adaptive procedure involves a two-stage sampling phase, starting with uniform allocation to eliminate suboptimal arms and estimate variances. This is followed by solving a Gaussian minimax game to determine a sampling policy and decision rule for the second stage. The strategy is proven to be simultaneously asymptotically minimax and Bayes optimal for simple regret, achieving upper bounds that match established lower bounds without requiring knowledge of outcome distributions or priors. AI
RANK_REASON The cluster contains a single academic paper published on arXiv. [lever_c_demoted from research: ic=1 ai=0.4]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
