DCMTL network: A double-contrast multi-task learning network for semi-supervised multi-source data classification

doi:10.1016/j.eswa.2025.129062

DCMTL network: A double-contrast multi-task learning network for semi-supervised multi-source data classification

2025-07-19

0

PRE

AI

DOI：10.1016/j.eswa.2025.129062

Original

Original request for help

Save

Abstract

En 中文

• A semi-supervised learning framework for multi-source data classification was proposed. • A contrastive learning approach considering instance discrimination, semantic clustering and multi-source index information was used. • An enhanced feature compactness strategy based on contrastive learning was implemented for better classification boundaries. • A multi-task learning method was performed for joint learning of multiple data sources.

Keywords:

semi-supervised learning

contrastive learning

multi-source data classification

feature compactness

multi-task learning

DCMTL network: A double-contrast multi-task learning network for semi-supervised multi-source data classification

Abstract

Journal

Organization