Basic belief assignment determination based on contrastive learning for pattern classification

Wei Li

doi:10.1016/j.inffus.2026.104572

Basic belief assignment determination based on contrastive learning for pattern classification

2026-06-25

0

PRE

AI

DOI：10.1016/j.inffus.2026.104572

Original

Original request for help

Save

Abstract

En 中文

• An Evidential Siamese Network (ESN) is proposed for BBA determination • ESN learns BBA determination end-to-end without predefined focal-element structures. • ESN uses contrastive learning to pull intra-class closer and inter-class apart. • Downstream task feedback is used to further improve the classification performance. • ESN outperforms prevailing BBA methods on multiple image and UCI datasets.

Keywords:

Dempster–Shafer evidence theory

Basic belief assignment

Uncertainty modeling

Contrastive learning

Pattern classification

Basic belief assignment determination based on contrastive learning for pattern classification

Abstract

Journal

Organization