Gaussian processes with prior-model-informed kernel for dynamical system modeling

Shengbing Tang

doi:10.1016/j.neunet.2026.109569

Gaussian processes with prior-model-informed kernel for dynamical system modeling

2026-09-01

0

PRE

AI

DOI：10.1016/j.neunet.2026.109569

Original

Original request for help

Save

Abstract

En 中文

• Introduces a novel kernel design informed by prior models. • Integrates physical models or simulators into the kernel. • Enhances GP extrapolation with well-calibrated uncertainty estimates. • Outperforms mean-based models in accuracy and uncertainty estimation. • Achieves superior results in active learning and reinforcement learning tasks.

Keywords:

Gaussian processes

Kernel design

Prior knowledge integration

Uncertainty estimation

Dynamical system modeling

Gaussian processes with prior-model-informed kernel for dynamical system modeling

Abstract

Journal

Organization

Cited Papers