Learning Physics with Neural Networks: Physics-Informed Models and Neural Operators for Scalable Scientific Computing
Seminar Announcement
Department of Fisheries Oceanography
"Learning Physics with Neural Networks: Physics-Informed Models and Neural Operators for Scalable Scientific Computing"
Elham Kianiharchegani, Postdoctoral Research Associate, Applied Mathematics Department, Brown University
Wednesday, February 11, 2026
3:00 - 4:00 pm
SMAST E 101-103 and via Zoom
Abstract:
Physics-informed neural networks (PINNs) offer a flexible framework for solving differential equations by embedding physical laws directly into neural network training. In this talk, I demonstrate how PINNs can be used as mesh-free solvers for both forward simulation and inverse problems, enabling parameter estimation, system identification, and data assimilation directly from sparse or noisy observations. Beyond classical formulations, I present several recent developments that improve the robustness and scalability of physics-informed learning. These include domain-decomposed approaches such as XPINNs for multi-region and multi-scale problems, second-order and quasi-Newton optimization strategies that significantly accelerate and stabilize training, and the integration of symbolic regression techniques for discovering governing equations directly from data. I will further discuss operator learning methods, including physics-informed DeepONets and related neural operator architectures, which directly learn mappings between input functions and their solutions, enabling fast and reusable surrogate models for entire families of parametric PDE problems. The talk highlights both methodological advances and practical applications in real-world engineering and physical systems, illustrating how physics-informed and operator-based models can complement and accelerate traditional numerical solvers within modern computational workflows.
Join Meeting
https://umassd.zoom.us/j/93758230260
Meeting ID and passcode required. Please email contact to obtain.
SMAST East 101-103
: 836 S. Rodney French Boulevard, New Bedford MA 02744
Callie Rumbut
c.rumbut@umassd.edu
https://umassd.zoom.us/j/93758230260