DeepONet as a Multi-Operator Extrapolation Model: Distributed Pretraining with Physics-Informed Fine-Tuning

doi:10.1016/j.jcp.2025.114537

DeepONet as a Multi-Operator Extrapolation Model: Distributed Pretraining with Physics-Informed Fine-Tuning

2025-11-19

0

PRE

AI

DOI：10.1016/j.jcp.2025.114537

Original

Original request for help

Save

Abstract

En 中文

• Proposed distributed pretraining and zero-shot physics-informed fine-tuning for multi operator learning. • Demonstrated effective strategies to improve operator learning generalization through fine tuning on pretrained models. • Combined distributed pretraining and low-rank adaptation to enable rapid adaptation. • Validated the approach with numerical examples, showing notable improvements in accu racy and efficacy.

DeepONet as a Multi-Operator Extrapolation Model: Distributed Pretraining with Physics-Informed Fine-Tuning

Abstract

Journal

Organization