Scientific Machine Learning: Principles and Applications

A recent report by the US Department of Energy defines the area of scientific machine learning as “a core component of artificial intelligence (AI) and a computational technology that can be trained, with scientific data, to augment or automate human skills”, which has “the potential to transform science and energy research”. In this presentation, we introduce supervised and unsupervised learning methods for scientific machine learning applied to microwave circuit and system design and simulation. We discuss the concepts of Physics-Informed Neural Networks (PINN) and Deep-Operator Networks that directly integrate physical laws into their loss function, so that the training process does not rely on the generation of training ground truth data from simulations (as in standard neural networks). We elucidate these concepts through four applications: electro-thermal modeling of MEMS with thermally-dependent materials, simulation of frequency selective surfaces (unsupervised learning), design optimization of reconfigurable intelligent surfaces and indoor radio coverage prediction (supervised learning).