Regularized-OFU: an efficient algorithm for general contextual bandit with optimization oracles
In contextual bandit, one major challenge is to develop theoretically solid and empirically efficient algorithms for general function classes. We present a novel algorithm called \emph{regularized optimism in face of uncertainty (ROFU)} for general contextual bandit problems. It exploits an optimization oracle to calculate the well-founded upper confidence bound (UCB). Theoretically, for general function classes under very mild assumptions, it achieves a near-optimal regret bound $\Tilde{O}(\sqrt{T})$. Practically, one great advantage of ROFU is that the optimization oracle can be efficiently implemented with low computational cost. Thus, we can easily extend ROFU for contextual bandits with deep neural networks as the function class, which outperforms strong baselines including the UCB and Thompson sampling variants.
Code (0)
등록된 구현이 없습니다.
Tasks
Multi-Armed BanditsThompson SamplingSimilar Papers 제목 키워드 기반
Regularized OFU: an Efficient UCB Estimator forNon-linear Contextual Bandit
Balancing exploration and exploitation (EE) is a fundamental problem in contex-tual bandit. One powerful principle for EE trade-off isOptimism in Face of Uncer-tainty(OFU), in which the agent takes the action according t…
Multi-Armed BanditsOracle-Efficient Pessimism: Offline Policy Optimization in Contextual Bandits
We consider offline policy optimization (OPO) in contextual bandits, where one is given a fixed dataset of logged interactions. While pessimistic regularizers are typically used to mitigate distribution shift, prior impl…
Multi-Armed BanditsOnline KL-Regularized Reinforcement Learning with Function Approximation under Misspecification
We study KL-regularized contextual bandits and episodic reinforcement learning (RL) under general function approximation with model misspecification. Existing guarantees rely on realizability and therefore do not extend …
Reinforcement LearningSurrogate Objectives for Batch Policy Optimization in One-step Decision Making
We investigate batch policy optimization for cost-sensitive classification and contextual bandits---two related tasks that obviate exploration but require generalizing from observed rewards to action selections in unseen…
Decision MakingMulti-Armed BanditsRegularized Contextual Bandits
We consider the stochastic contextual bandit problem with additional regularization. The motivation comes from problems where the policy of the agent must be close to some baseline policy which is known to perform well o…
Multi-Armed Bandits