Handling unstructured data for operator learning using implicit neural representations

Thomas X Wang, patrick gallinari

Keywords: Operator learning, implicit neural representations

TL;DR: We propose a novel method to extend operator learning methods for unstructured data.

Abstract: Operator learning methods are too often constrained by a fixed sampling of both the input and output functions. We propose a novel method to allow current operator learning methods to learn on any sampling. We show that our method can perform inference on unseen samplings, and that it allows returning outputs as continuous functions.