RDGCL: Reaction-Diffusion Graph Contrastive Learning for Recommendation

  • 2024-08-22 13:50:09
  • Jeongwhan Choi, Hyowon Wi, Chaejeong Lee, Sung-Bae Cho, Dongha Lee, Noseong Park
  • 0

Abstract

Contrastive learning (CL) has emerged as a promising technique for improvingrecommender systems, addressing the challenge of data sparsity by usingself-supervised signals from raw data. Integration of CL with graphconvolutional network (GCN)-based collaborative filterings (CFs) has beenexplored in recommender systems. However, current CL-based recommendationmodels heavily rely on low-pass filters and graph augmentations. In this paper,inspired by the reaction-diffusion equation, we propose a novel CL method forrecommender systems called the reaction-diffusion graph contrastive learningmodel (RDGCL). We design our own GCN for CF based on the equations ofdiffusion, i.e., low-pass filter, and reaction, i.e., high-pass filter. Ourproposed CL-based training occurs between reaction and diffusion-basedembeddings, so there is no need for graph augmentations. Experimentalevaluation on 5 benchmark datasets demonstrates that our proposed methodoutperforms state-of-the-art CL-based recommendation models. By enhancingrecommendation accuracy and diversity, our method brings an advancement in CLfor recommender systems.

 

Quick Read (beta)

loading the full paper ...