← All Research Themes
Neuro-Symbolic Knowledge Discovery and Multi-Physical Simulation
Overview
Using symbolic discovery, neural-symbolic methods, and physics-aware models to uncover governing structure and simulate complex coupled systems.
Key Questions
- How can symbolic regression discover compact, interpretable governing laws from noisy data?
- How can learned models respect physical constraints while remaining expressive enough for coupled systems?
- How can we combine symbolic discovery, neural operators, and simulation to support multiphysics modeling?
Our Contributions
- Developed symbolic regression methods for PDEs and structure-aware scientific modeling
- Explored physics-preserving operators and neural-symbolic strategies for multiphysics problems
- Designed hybrid frameworks that connect symbolic discovery with simulation-oriented learning
Recommended Papers
Prunes symbolic regressors with physics residuals and sensitivity, yielding cleaner equations.
Uses T-NNGP to produce interpretable multi-physics PDE solutions.
Embeds physics in graph networks for stable long-horizon multiphysics simulation.
Recent Work
Open Problems
- Scaling symbolic discovery to larger coupled PDE systems and multiphysics settings
- Handling sparse, noisy, and heterogeneous scientific observations
- Maintaining interpretability while improving robustness and generalization
Prospective Student Projects
- Symbolic regression for PDE discovery
- Physics-aware neural operators for coupled systems
- Hybrid symbolic–numeric simulation frameworks