← 研踪 PaperTrailarXiv:2606.11171 · v11
Bellman-sufficient Information Complexity
首次提交 2026-06-10 01:49
版本更新 2026-11-16 07:29
雷达观测 2026-11-17 17:12
摘要
We introduce Bellman-sufficient information complexity for minimax analysis of sequential decision problems. A Bellman-sufficient state retains the history needed for the controlled recursion; an index specifies which information the bounds charge. A log-penalized Bellman program gives the upper bound, and a Bellman--Fano comparison with an algorithm-dependent reference trajectory gives the lower bound. When admissibility, calibration, and growth conditions hold and the bounds match at a common localization scale, they yield an information-risk sandwich. UCB bounds the immediate decision loss through calibration and optimism, while AMS/EBO gives a robust relaxation of the AIR Bellman program. Our main application gives a negative answer to a widely studied form of the GP--UCB minimax-optimality question. For every , we construct a fixed bounded continuous kernel with minimax regret along an infinite sequence of horizons. At those horizons, two globally calibrated GP--UCB rules incur linear regret under one fixed truth. A policy that applies action-index AMS to a predetermined finite set in each epoch attains the minimax order; EBO gives a dual interpretation. Many low-value directions inflate the global exploration multiplier even though ignoring them is inexpensive. Through the canonical RKHS feature map, the same construction gives a finite-horizon polynomial minimax gap for the specified maximal-information-calibrated LinUCB rule. A numerical experiment illustrates the mechanism.
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v11 · 2026-11-16 07:29 · 最新观测版本
Bellman-sufficient Information Complexity
We introduce Bellman-sufficient information complexity for minimax analysis of sequential decision problems. A Bellman-sufficient state retains the history needed for the controlled recursion; an index specifies which information the bounds charge. A log-penalized Bellman program gives the upper bound, and a Bellman--Fano comparison with an algorithm-dependent reference trajectory gives the lower bound. When admissibility, calibration, and growth conditions hold and the bounds match at a common localization scale, they yield an information-risk sandwich. UCB bounds the immediate decision loss through calibration and optimism, while AMS/EBO gives a robust relaxation of the AIR Bellman program. Our main application gives a negative answer to a widely studied form of the GP--UCB minimax-optimality question. For every , we construct a fixed bounded continuous kernel with minimax regret along an infinite sequence of horizons. At those horizons, two globally calibrated GP--UCB rules incur linear regret under one fixed truth. A policy that applies action-index AMS to a predetermined finite set in each epoch attains the minimax order; EBO gives a dual interpretation. Many low-value directions inflate the global exploration multiplier even though ignoring them is inexpensive. Through the canonical RKHS feature map, the same construction gives a finite-horizon polynomial minimax gap for the specified maximal-information-calibrated LinUCB rule. A numerical experiment illustrates the mechanism.
查看此版本解读 →查看此版本原文 ↗