Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Contrastive Learning Augmented Social Recommendations

Lin Wang, Weisong Wang, Xuanji Xiao, Qing Li

TL;DR

The paper addresses cold-start challenges in recommender systems by fusing social-relations with behavior data. It introduces CLSRec, which denoises the social graph via a low-rank SVD reconstruction and aligns it with the original view through contrastive learning, then disentangles aligned versus domain-specific interests with mutual distillation and co-attention, finally fusing signals with gated routing. Key contributions include the $A^{S} = U_k S_k U_k^{\top}$ social reconstruction, the contrastive loss $\mathcal{L}_{cl}$, and the mutual distillation framework that yields both shared and specific representations. Empirical results on LastFM and Ciao show clear improvements, especially for cold users, highlighting the practical potential of robust social-behavior fusion in recommendation systems.

Abstract

Recommender systems are essential for modern content platforms, yet traditional behavior-based models often struggle with cold users who have limited interaction data. Engaging these users is crucial for platform growth. To bridge this gap, we propose leveraging the social-relation graph to enrich interest representations from behavior-based models. However, extracting value from social graphs is challenging due to relation noise and cross-domain inconsistency. To address the noise propagation and obtain accurate social interest, we employ a dual-view denoising strategy, employing low-rank SVD to the user-item interaction matrix for a denoised social graph and contrastive learning to align the original and reconstructed social graphs. Addressing the interest inconsistency between social and behavioral interests, we adopt a "mutual distillation" technique to isolate the original interests into aligned social/behavior interests and social/behavior specific interests, maximizing the utility of both. Experimental results on widely adopted industry datasets verify the method's effectiveness, particularly for cold users, offering a fresh perspective for future research. The implementation can be accessed at https://github.com/WANGLin0126/CLSRec.

Contrastive Learning Augmented Social Recommendations

TL;DR

The paper addresses cold-start challenges in recommender systems by fusing social-relations with behavior data. It introduces CLSRec, which denoises the social graph via a low-rank SVD reconstruction and aligns it with the original view through contrastive learning, then disentangles aligned versus domain-specific interests with mutual distillation and co-attention, finally fusing signals with gated routing. Key contributions include the social reconstruction, the contrastive loss , and the mutual distillation framework that yields both shared and specific representations. Empirical results on LastFM and Ciao show clear improvements, especially for cold users, highlighting the practical potential of robust social-behavior fusion in recommendation systems.

Abstract

Recommender systems are essential for modern content platforms, yet traditional behavior-based models often struggle with cold users who have limited interaction data. Engaging these users is crucial for platform growth. To bridge this gap, we propose leveraging the social-relation graph to enrich interest representations from behavior-based models. However, extracting value from social graphs is challenging due to relation noise and cross-domain inconsistency. To address the noise propagation and obtain accurate social interest, we employ a dual-view denoising strategy, employing low-rank SVD to the user-item interaction matrix for a denoised social graph and contrastive learning to align the original and reconstructed social graphs. Addressing the interest inconsistency between social and behavioral interests, we adopt a "mutual distillation" technique to isolate the original interests into aligned social/behavior interests and social/behavior specific interests, maximizing the utility of both. Experimental results on widely adopted industry datasets verify the method's effectiveness, particularly for cold users, offering a fresh perspective for future research. The implementation can be accessed at https://github.com/WANGLin0126/CLSRec.

Paper Structure

This paper contains 12 sections, 10 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Information Extraction of Dual-view
  4. Refine the Social knowledge
  5. Reconstructed Social View
  6. Contrastive Learning Alignment
  7. Interest Inconsistency Alignment
  8. Training and Optimization
  9. Experiments
  10. Performance Comparison
  11. Ablation Study
  12. Conclusion

Figures (3)

  • Figure 1: Social interest information on a typical video platform encompasses: (1) tracking stars; (2) keeping up with uploaders; (3) subscribing to TV series.
  • Figure 2: Our proposed contrastive learning social recommendation (CLSRec) framework.
  • Figure 3: Interest-inconsistency alignment by mutual distillation and dynamic routing.