Asynchronous parallel surrogate optimization aided neural network design with variable evaluation runtime for streamflow and pollutant forecast

doi:10.1016/j.jhydrol.2025.134378

Asynchronous parallel surrogate optimization aided neural network design with variable evaluation runtime for streamflow and pollutant forecast

2025-10-08

0

PRE

AI

Christine A. Shoemaker

DOI：10.1016/j.jhydrol.2025.134378

Original

Original request for help

Save

Abstract

En 中文

• Hyperparameter optimization (HPO) in hydrological forecasting is highly multi-modal and evaluation runtime varied. • A new asynchronous parallel surrogate global optimization method, ASONN, for HPO of neural network models. • ASONN accelerates HPO by up to 60% compared to asynchronous parallel global optimization method.

Asynchronous parallel surrogate optimization aided neural network design with variable evaluation runtime for streamflow and pollutant forecast

Abstract

Journal

Organization