← 研踪 PaperTrailarXiv:2603.28595 · v3
Optimistic Actor-Critic with Parametric Policies for Linear Markov Decision Processes
首次提交 2026-03-30 23:41
版本更新 2026-11-18 18:42
雷达观测 2026-11-19 15:04
摘要
Although actor-critic methods have been successful in practice, their theoretical analyses have several limitations. Specifically, existing theoretical work either sidesteps the exploration problem by making strong assumptions or analyzes impractical methods with complicated algorithmic modifications. Moreover, the actor-critic methods analyzed for linear MDPs often employ natural policy gradient and define "implicit" policies through the accumulated critic estimates, rather than learning an explicitly parameterized actor. Such policies can be computationally expensive to sample from, making the environment interactions inefficient. To that end, we focus on the finite-horizon linear MDPs and propose an optimistic actor-critic framework that uses parametric log-linear policies. In particular, we introduce a tractable logit-matching regression objective for the actor. For the critic, we use approximate Thompson sampling via Langevin Monte Carlo to obtain optimistic value estimates. We prove that, when the policy features are sufficiently expressive, the resulting algorithm achieves and sample complexity in the on-policy and off-policy settings, respectively. Our results match the state-of-the-art dependence on in prior theoretical work, while our algorithm is more aligned with practice.
作者备注:AISTATS 2026
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。
v3 · 2026-11-18 18:42 · 最新观测版本
Optimistic Actor-Critic with Parametric Policies for Linear Markov Decision Processes
Although actor-critic methods have been successful in practice, their theoretical analyses have several limitations. Specifically, existing theoretical work either sidesteps the exploration problem by making strong assumptions or analyzes impractical methods with complicated algorithmic modifications. Moreover, the actor-critic methods analyzed for linear MDPs often employ natural policy gradient and define "implicit" policies through the accumulated critic estimates, rather than learning an explicitly parameterized actor. Such policies can be computationally expensive to sample from, making the environment interactions inefficient. To that end, we focus on the finite-horizon linear MDPs and propose an optimistic actor-critic framework that uses parametric log-linear policies. In particular, we introduce a tractable logit-matching regression objective for the actor. For the critic, we use approximate Thompson sampling via Langevin Monte Carlo to obtain optimistic value estimates. We prove that, when the policy features are sufficiently expressive, the resulting algorithm achieves and sample complexity in the on-policy and off-policy settings, respectively. Our results match the state-of-the-art dependence on in prior theoretical work, while our algorithm is more aligned with practice.
查看此版本解读 →查看此版本原文 ↗