PHYS brown bag seminar: "Physics-Informed Neural Networks with NVIDIA Modulus"

Abstract:

NVIDIA Modulus is a Pytorch-based open-source framework for solving science and engineering problems using AI. Modulus has been successfully applied to solve a wide range of problems from structural mechanics to fluid dynamics. In this entry-level talk, we will review the fundamentals of Physics-informed Neural Networks (PINNs) and present example cases that cover parameterized problems, inverse problems and time-dependent problems.

Physics-informed neural networks denote neural networks that are trained by constraining the loss function to satisfy governing equations e.g. a system of PDEs. Physics-informed neural networks can also be combined with a data-driven loss term to accelerate training or to invert out unknown system variables. They allow the solution of several instances of the PDEs at once by parameterizing the input space e.g. initial and boundary conditions and geometry. Therefore, the model training step can
be undertaken only once using massively parallel modern GPU-architectures and subsequently the parameter space exploration can be performed in near-real time via inference.

Speaker's bio:

Niki is an AI/HPC solutions architect at NVIDIA Helsinki, helping academic researchers to leverage NVIDIA’s technology stack through the NVIDIA AI Technology Center program. Prior to joining NVIDIA, he worked as a researcher in the Department of Aeronautics at Imperial College London, where he also obtained his PhD in Computational Fluid Dynamics. Specifically, his research focused on the development of high-order accurate numerical methods for solving large-scale incompressible fluid
flow problems using massively parallel modern GPU architectures. Currently, Niki is particularly interested in physics-informed neural networks and scientific ML in general.