AI RESEARCH

Functional-prior-based Bayesian PDE-constrained inversion using PINNs

arXiv CS.LG

ArXi:2605.07060v1 Announce Type: cross Physics-informed neural networks (PINNs) provide a mesh-free framework for solving PDE-constrained inverse problems, but their extension to Bayesian inversion still faces a fundamental difficulty: prior distributions are typically defined in the weight space of neural networks, whereas physically meaningful prior assumptions are naturally expressed in function space. In this study, we