arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2602.17832cs.LGcs.RO
收藏

MePoly:最大熵多项式策略优化

MePoly: Max Entropy Polynomial Policy Optimization

University of Michigan, Ann Arbor(密歇根大学,安阿伯分校);University of California, Berkeley(加州大学伯克利分校)

Hang Liu, Sangli Teng, Maani Ghaffari

AI总结:

MePoly通过多项式能量模型实现最大熵策略优化,有效捕捉复杂非凸流形并在多种基准上优于基线方法。

摘要:

随机最优控制提供了一个统一的数学框架,用于解决复杂决策问题,涵盖了最大熵强化学习(RL)和模仿学习(IL)等范式。然而,传统参数化策略往往难以表示解的多模态性。尽管基于扩散的策略旨在恢复多模态性,但它们缺乏显式的概率密度,这使得策略梯度优化变得复杂。为弥合这一差距,我们提出了MePoly,一种基于多项式能量模型的新型策略参数化方法。MePoly提供了一个显式且可计算的概率密度,使精确熵最大化成为可能。理论上,我们基于经典的矩问题,利用其对任意分布的通用逼近能力来支撑我们的方法。经验上,我们证明MePoly能够有效捕捉复杂的非凸流形,并在多样本基准上优于基线方法。

英文摘要:

Stochastic Optimal Control provides a unified mathematical framework for solving complex decision-making problems, encompassing paradigms such as maximum entropy reinforcement learning(RL) and imitation learning(IL). However, conventional parametric policies often struggle to represent the multi-modality of the solutions. Though diffusion-based policies are aimed at recovering the multi-modality, they lack an explicit probability density, which complicates policy-gradient optimization. To bridge this gap, we propose MePoly, a novel policy parameterization based on polynomial energy-based models. MePoly provides an explicit, tractable probability density, enabling exact entropy maximization. Theoretically, we ground our method in the classical moment problem, leveraging the universal approximation capabilities for arbitrary distributions. Empirically, we demonstrate that MePoly effectively captures complex non-convex manifolds and outperforms baselines in performance across diverse benchmarks.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑