Negative-weighted knowledge distillation regularized graph convolutional network for multi-label class-incremental learning

Kaile Du

doi:10.1016/j.patcog.2026.113657

Negative-weighted knowledge distillation regularized graph convolutional network for multi-label class-incremental learning

2026-04-06

0

PRE

AI

DOI：10.1016/j.patcog.2026.113657

Original

Original request for help

Save

Abstract

En 中文

• Asymmetric label absence is defined and analyzed in MLCIL. • A CGCN dynamically builds reliable cross-task label relationships. • NKD corrects label imbalance and reduces false positive predictions. • The N-CGCN achieves state-of-the-art results on MS-COCO and VOC 2007.

Keywords:

Multi-label class-incremental learning

Graph convolutional network

Label imbalance

Knowledge distillation

Cross-task label relationships

Negative-weighted knowledge distillation regularized graph convolutional network for multi-label class-incremental learning

Abstract

Journal

Organization

Cited Papers