A graph contrastive learning framework with multi-view adaptation for robust collaborative filtering

Dongyu Jiang

doi:10.1016/j.asoc.2026.115358

A graph contrastive learning framework with multi-view adaptation for robust collaborative filtering

2026-04-29

0

PRE

AI

DOI：10.1016/j.asoc.2026.115358

Original

Original request for help

Save

Abstract

En 中文

• MAGCL integrates multi-view contrastive learning into graph-based collaborative filtering. • A high-order collaborative graph effectively captures richer interaction signals in sparse data. • A trust score-based denoising mechanism filters unreliable interactions, enhancing robustness. • Dynamic view fusion and cross-layer contrastive loss significantly improve generalization. • A uniformity loss optimizes embedding quality, ensuring balanced representation distribution.

Keywords:

graph contrastive learning

multi-view adaptation

collaborative filtering

robustness

embedding quality

A graph contrastive learning framework with multi-view adaptation for robust collaborative filtering

Abstract

Journal

Organization