Cumulative Regret
A metric that measures the cumulative loss of decisions made at each step during an optimization search process. It focuses on the performance of the entire process, in contrast to simple regret.
Timeline (last 90 days)
The multi-armed bandit (MAB) is a classic framework in reinforcement learning and online decision-making, in which a player faces K arms, each arm's reward follows an unknown distribution, and the goal is to minimize cumulative regret.
This study proposes that the existing cumulative regret guarantees of Bayesian optimization do not necessarily characterize the quality of the algorithm's final output.
The cumulative regret upper bound of standard GP-UCB is parameterized by the RKHS norm bound and the maximum information gain.
All Facts (3)
The multi-armed bandit (MAB) is a classic framework in reinforcement learning and online decision-making, in which a player faces K arms, each arm's reward follows an unknown distribution, and the goal is to minimize cumulative regret.
50%UnverifiedThis study proposes that the existing cumulative regret guarantees of Bayesian optimization do not necessarily characterize the quality of the algorithm's final output.
50%UnverifiedThe cumulative regret upper bound of standard GP-UCB is parameterized by the RKHS norm bound and the maximum information gain.
50%