One Operator to Rule Them All? On Boundary-Indexed Operator Families in Neural PDE Solvers

TL;DR

Neural PDE solvers implicitly learn boundary-conditioned operator families, which limits their generalization across different boundary conditions and highlights the need for explicit boundary-aware modeling.

Contribution

This work formalizes the boundary-conditioning of neural operators and demonstrates its impact on generalization, revealing a core limitation of current neural PDE solvers.

Findings

Sharp degradation under boundary-condition shifts

Failures in cross-distribution boundary ensembles

Convergence to boundary-conditioned expectations when boundary info is removed

Abstract

Neural PDE solvers are often described as learning solution operators that map problem data to PDE solutions. In this work, we argue that this interpretation is generally incorrect when boundary conditions vary. We show that standard neural operator training implicitly learns a boundary-indexed family of operators, rather than a single boundary-agnostic operator, with the learned mapping fundamentally conditioned on the boundary-condition distribution seen during training. We formalize this perspective by framing operator learning as conditional risk minimization over boundary conditions, which leads to a non-identifiability result outside the support of the training boundary distribution. As a consequence, generalization in forcing terms or resolution does not imply generalization across boundary conditions. We support our theoretical analysis with controlled experiments on the Poisson equation, demonstrating sharp degradation under boundary-condition shifts, cross-distribution failures between distinct boundary ensembles, and convergence to conditional expectations when boundary information is removed. Our results clarify a core limitation of current neural PDE solvers and highlight the need for explicit boundary-aware modeling in the pursuit of foundation models for PDEs.