A continually evolving knowledge-guided deep learning framework for daily maize yield formation

Junji Ou

doi:10.1016/j.agsy.2026.104981

A continually evolving knowledge-guided deep learning framework for daily maize yield formation

2026-09-14

0

PRE

AI

William David Batchelor

DOI：10.1016/j.agsy.2026.104981

Original

Original request for help

Save

Abstract

En 中文

• A knowledge-guided framework simulates grain growth processes for yield prediction. • Physiological processes are explicitly represented to ensure interpretability. • Yield and grain number errors are reduced by 24% and 29% relative to APSIM. • Key physiological information is preserved during model evolving. • Expert-constrained training reduces the need for manual parameter calibration.

Keywords:

Knowledge-guided deep learning

Continual learning

Daily yield formation simulation

Hybrid modeling

Physiological interpretability

A continually evolving knowledge-guided deep learning framework for daily maize yield formation

Abstract

Journal

Organization