[Deep Learning Theory Team Seminar] Talk by Prof. Wuyang Chen (SFU) on Building Machines That Understand the Physics

Title: Building Machines That Understand the Physics

Abstract:
In this talk, we explore how to build cutting-edge AI models that understand and reason about physical systems. Our approach integrates two key strategies. First, Tool Interaction: Rather than embedding physical knowledge directly into data-driven models, we teach AI models to interact with scientific tools. Second, Physics-Enriched Data Scaling: Instead of merely increasing the volume of training samples, we make our data more physically informative by refining it with scientific tools.
We illustrate our methodology with two showcase applications. First, to support multi-step scientific reasoning, we autoformalize and generate code for PDE solvers, and align large language models (LLMs) with simulation feedback via preference optimization. Second, to enable real-time interactive fluid modeling, we hybridize neural physics with traditional numerical solvers, and assist users' freehand sketches by learning to generate external force fields through reverse simulation.

Dr. Wuyang Chen is a tenure-track Assistant Professor in Computing Science at Simon Fraser University. Previously, he was a postdoctoral researcher in Statistics at the University of California, Berkeley, advised by Professor Michael Mahoney. He obtained his Ph.D. in Electrical and Computer Engineering from the University of Texas at Austin in 2023, advised by Professor Atlas Wang. Dr. Chen's research focuses on integrating AI methods with physical knowledge, scientific machine learning, and theoretical understanding of deep networks. Dr. Chen has published papers at CVPR, ECCV, ICLR, ICML, NeurIPS, and other top conferences. Dr. Chen's research has been recognized by NSF (National Science Foundation) newsletter in 2022, INNS Doctoral Dissertation Award and the iSchools Doctoral Dissertation Award in 2024, and AAAI New Faculty Highlights in 2025. Dr. Chen is the host of the Foundation Models for Science workshop at NeurIPS 2024 and co-organized the 4th and 5th versions of the UG2+ workshop and challenge at CVPR in 2021 and 2022.