Description
Partial Differential Equations (PDEs) are ubiquitous as mathematical models in science and engineering. Numerical methods are the key tools in their study. Yet, despite their considerable success, the underlying high computational cost inhibits the scope and applicability of numerical methods and necessitates the development of alternatives. In this context, machine learning (ML) algorithms are being increasingly used as surrogates for numerical PDE solvers. In this talk, We will focus on a particular class of ML algorithms that aim to learn the solution operator of PDEs from data. The underlying issues will discussed and different architectures presented. In particular, We will present a very recently proposed Foundation Model for PDEs.
