SGCL: Unifying Self-Supervised and Supervised Learning for Graph Recommendation

TL;DR

This paper introduces SGCL, a unified framework that combines supervised and self-supervised learning for graph-based recommendation, resulting in faster training and improved accuracy over existing methods.

Contribution

SGCL is the first to unify supervised and self-supervised graph learning into a single contrastive framework for recommendation.

Findings

SGCL outperforms state-of-the-art methods in accuracy.

SGCL achieves faster training times.

SGCL demonstrates superior efficiency and effectiveness.

Abstract

Recommender systems (RecSys) are essential for online platforms, providing personalized suggestions to users within a vast sea of information. Self-supervised graph learning seeks to harness high-order collaborative filtering signals through unsupervised augmentation on the user-item bipartite graph, primarily leveraging a multi-task learning framework that includes both supervised recommendation loss and self-supervised contrastive loss. However, this separate design introduces additional graph convolution processes and creates inconsistencies in gradient directions due to disparate losses, resulting in prolonged training times and sub-optimal performance. In this study, we introduce a unified framework of Supervised Graph Contrastive Learning for recommendation (SGCL) to address these issues. SGCL uniquely combines the training of recommendation and unsupervised contrastive losses into a cohesive supervised contrastive learning loss, aligning both tasks within a single optimization direction for exceptionally fast training. Extensive experiments on three real-world datasets show that SGCL outperforms state-of-the-art methods, achieving superior accuracy and efficiency.