题   目：AI-Assisted Discovery of Symbolic Laws and Decision Formulas

嘉　宾：周源 副教授 清华大学

主持人：曾智宇 助理教授 上海交通大学安泰经济与管理学院

时　间：2026年9月11日（周五）14:00-15:30

地　点：上海交通大学安泰经济与管理学院包兆龙图书馆B207

内容简介：

Discovering concise and interpretable mathematical laws from observational data while respecting domain knowledge is a central problem in AI-assisted scientific discovery.

The first part of this talk presents Physical End-to-End Symbolic Regression (PhyE2E). Symbolic regression seeks to discover mathematical formulas that directly characterize relationships among variables. However, it faces several fundamental challenges, including the exponential growth of the search space with expression complexity, sparse reward signals, the difficulty of incorporating physical priors, and the trade-off between predictive accuracy and formula simplicity. PhyE2E integrates physical dimensions, formula complexity, candidate operators, and constants into an end-to-end model. It learns dimensional consistency while generating formulas, decomposes complex multivariate expressions into smaller subproblems through variable splitting, and employs search algorithms for local refinement. Experiments demonstrate substantial advantages in symbolic accuracy, dimensional accuracy, and formula complexity. We further apply PhyE2E to several problems in space physics. The resulting formulas characterize both short- and long-term cycles of sunspot activity, reveal a physically interpretable relationship governing near-Earth plasma pressure, and describe differential solar rotation in high-latitude regions where observations are limited. These studies illustrate how AI can move beyond fitting observational data to assist scientists in discovering mathematical expressions that can be interpreted, tested, and investigated further.

The second part introduces ongoing work that applies symbolic search to operational problems to automatically discover simple and transparent decision policies. High-quality policies for complex operational problems typically depend on model-specific analysis and manual derivation. Policies learned by neural networks can address complex problems but are often difficult to interpret and rarely admit rigorous performance guarantees. To address this challenge, we explore symbolic approaches to operational decision-making that use AI to discover concise, interpretable, and directly implementable decision formulas. We further investigate AI-assisted methods for analyzing their theoretical properties, with the goal of establishing provable guarantees for AI-discovered decision rules.

演讲人简介：

Yuan Zhou is an Associate Professor at the Yau Mathematical Sciences Center, Tsinghua University. He received his B.Eng. in Computer Science from Tsinghua University in 2009 and his Ph.D. in Computer Science from Carnegie Mellon University in 2014. Prior to joining Tsinghua, he was an Instructor in Applied Mathematics at the Massachusetts Institute of Technology and an Assistant Professor at the University of Illinois Urbana-Champaign and Indiana University Bloomington.

His research focuses on data-driven decision-making and artificial intelligence for science, spanning operations research, machine learning, and optimization. His work has appeared in leading journals and conferences in operations research, management science, machine learning, and theoretical computer science, including Operations Research, Management Science, Mathematics of Operations Research, Production and Operations Management, SIAM Journal on Optimization, Nature Machine Intelligence, Journal of Machine Learning Research, ICML, NeurIPS, ICLR, COLT, STOC, FOCS, and SODA. He currently serves as an Associate Editor for Operations Research and Operations Research Letters.

欢迎广大师生参加！