Triple-Supervised Progressive Contrastive Learning for Heterogeneous Graph Embedding

doi:10.1016/j.inffus.2025.104051

Triple-Supervised Progressive Contrastive Learning for Heterogeneous Graph Embedding

2025-12-15

0

PRE

AI

DOI：10.1016/j.inffus.2025.104051

Original

Original request for help

Save

Abstract

En 中文

• A novel heterogeneous graph contrastive learning method is proposed to improve the quality of node representations. • A progressive contrastive strategy is developed with dynamic negative sampling to enhance discrimination. • A meta-path-based modeling module is introduced to effectively capture complex heterogeneous semantics. • Experiments on three public datasets are conducted to validate the performance of the proposed model.

Triple-Supervised Progressive Contrastive Learning for Heterogeneous Graph Embedding

Abstract

Journal

Organization