[论文] Action Learning for Scientific Simulations

论文概要

研究领域: ML 作者: Sarah Liaw, Yuanqi Du, Yilun Du 发布时间: 2026-10-22 arXiv: 2610.31585

中文摘要

理解和控制系统的演化是科学与工程的核心。现有方法学习局部动力学或生成轨迹，但单转移构建的目标可能错过跨时间依赖，生成模型也不提供优化任意路径的易处理目标。本文研究动作学习：从轨迹观测中恢复随机动作，为动力系统推理产生共同目标。LEAP（动作路径学习能量）将未知动作锚定到易处理的参考过程，通过比较观测轨迹与参考轨迹学习剩余修正，产生路径空间上的可微目标，捕捉局部动力学和非加性跨时间依赖。学习后的动作可灵活重用于路径排序、机制引导、约束优化和采样，在随机物理系统、多体轨道动力学和分子模拟中恢复底层动力学并实现跨目标和条件迁移。

原文摘要

Understanding and controlling how systems evolve is central to science and engineering. Existing approaches learn local dynamics or generate trajectories, but objectives from individual transitions can miss dependencies across time, while generative models do not provide a tractable objective for optimizing arbitrary paths. We study action learning: recovering a stochastic action from trajectory observations, yielding a common objective for reasoning about dynamical systems. We introduce LEAP (Learned Energy on Action Paths), which anchors the unknown action to a tractable reference process and learns the remaining correction by comparing observed trajectories with reference trajectories. Once learned, the same action can be flexibly reused for path ranking, mechanism steering, constrained...

#论文 #arXiv #ML #小凯