Tutorial 3
Sourajeet Roy, Indian Institute of Technology (IIT) Roorkee
Abstract -
Artificial neural networks (ANNs) have long been used as data-driven surrogate models for high-speed devices and circuits, enabling analyses at a fraction of the computational cost of traditional electromagnetic (EM) solvers. Despite their success in applications such as design optimization and uncertainty quantification, conventional ANNs suffer from high training costs, overfitting, limited extrapolation capability, and the inability to enforce the underlying laws of physics.
This tutorial introduces the emerging paradigm of physics-informed neural networks (PINNs), which embed governing PDEs and ODEs of target circuits directly into the training process. As a result, PINNs have shifted the state-of-the-art in artificial intelligence from data-based to physics-based techniques. In this tutorial, I will present the mathematical foundations of PINNs and demonstrate how they enable rapid, data-free training while producing physically consistent models. Using examples from signal and power integrity, I will highlight the advantages of PINNs over conventional ANNs, discuss their current limitations, and present recent advances from our Computational Modeling and Simulation laboratory at IIT Roorkee that address these challenges.
Prof. Sourajeet Roy (Senior Member, IEEE) received the B.Tech. degree in electrical engineering from Sikkim Manipal University, Gangtok, India, in 2006, and the M.E.Sc. and Ph.D. degrees in electrical engineering from the University of Western Ontario, London, ON, Canada, in 2009 and 2013, respectively. From 2019 to 2023, he was an Assistant Professor at the Department of Electronics and Communications Engineering, IIT Roorkee, Roorkee, India, where he led the Computational Modeling and Simulation (CMAS) Research Group. In 2023, he was promoted to the position of an Associate Professor at the Department of Electronics and Communications Engineering. Dr. Roy is a member of the Technical Committee on Electrical Design, Modeling, and Simulation of the IEEE Electronic Packaging Society. He served as the General Chair for the 17th IEEE Electrical Design of Advanced Packaging and Systems (EDAPS) Symposium. He also serves as an Associate Editor for IEEE TRANSACTIONS ON COMPONENTS, PACKAGING, AND MANUFACTURING TECHNOLOGY and IEEE JOURNAL ON MULTISCALE AND MULTIPHYSICS COMPUTATIONAL TECHNIQUES.