Coralia Cartis

An Optimisation Framework for the Well-Conditioned Training of Physics-Informed Neural Networks

Joseph Webb, Sadok Jerad, Coralia Cartis

Physics-informed neural networks (PINNs) have emerged as a promising route to solve partial differential equations, yet they have struggled to reach the precision of classical solvers. The obstacle is increasingly understood to be one of optimisation, owing to the severely ill-co…Read more