Abstract:Contrastive Learning (CL) has achieved impressive performance in self-supervised learning tasks, showing superior generalization ability. Inspired by the success, adopting CL into collaborative filtering (CF) is prevailing in semi-supervised top-K recommendations. The basic idea is to routinely conduct heuristic-based data augmentation and apply contrastive losses (e.g., InfoNCE) on the augmented views. Yet, some CF-tailored challenges make this adoption suboptimal, such as the issue of out-of-distribution, the risk of false negatives, and the nature of top-K evaluation. They necessitate the CL-based CF scheme to focus more on mining hard negatives and distinguishing false negatives from the vast unlabeled user-item interactions, for informative contrast signals. Worse still, there is limited understanding of contrastive loss in CF methods, especially w.r.t. its generalization ability. To bridge the gap, we delve into the reasons underpinning the success of contrastive loss in CF, and propose a principled Adversarial InfoNCE loss (AdvInfoNCE), which is a variant of InfoNCE, specially tailored for CF methods. AdvInfoNCE adaptively explores and assigns hardness to each negative instance in an adversarial fashion and further utilizes a fine-grained hardness-aware ranking criterion to empower the recommender's generalization ability. Training CF models with AdvInfoNCE, we validate the effectiveness of AdvInfoNCE on both synthetic and real-world benchmark datasets, thus showing its generalization ability to mitigate out-of-distribution problems. Given the theoretical guarantees and empirical superiority of AdvInfoNCE over most contrastive loss functions, we advocate its adoption as a standard loss in recommender systems, particularly for the out-of-distribution tasks. Codes are available at <a class="link-external link-https" href="https://github.com/LehengTHU/AdvInfoNCE" rel="external noopener nofollow">this https URL</a>.

Revisiting Recommendation Loss Functions through Contrastive Learning (Technical Report)

Personalized Representation with Contrastive Loss for Recommendation Systems

Debiased Contrastive Loss for Collaborative Filtering.

Empowering Collaborative Filtering with Principled Adversarial Contrastive Loss

Ranking-based Contrastive Loss for Recommendation Systems

Contrastive Learning for Recommender System

Ranking Info Noise Contrastive Estimation: Boosting Contrastive Learning via Ranked Positives

Contrastive Learning for Debiased Candidate Generation in Large-Scale Recommender Systems

SINCERE: Supervised Information Noise-Contrastive Estimation REvisited

Meta-optimized Contrastive Learning for Sequential Recommendation

InfoNCE: Identifying the Gap Between Theory and Practice

Graph Contrastive Learning with Hybrid Noise Augmentation for Recommendation.

Incorporating Bias-aware Margins into Contrastive Loss for Collaborative Filtering

Debiased Contrastive Representation Learning for Mitigating Dual Biases in Recommender Systems

DCIB: Dual contrastive information bottleneck for knowledge-aware recommendation

Neighborhood-Enhanced Supervised Contrastive Learning for Collaborative Filtering

Robust Contrastive Learning against Noisy Views

Model-Aware Contrastive Learning: Towards Escaping the Dilemmas

PopDCL: Popularity-aware Debiased Contrastive Loss for Collaborative Filtering

Contrastive Learning of Preferences with a Contextual InfoNCE Loss

TDCGL: Two-Level Debiased Contrastive Graph Learning for Recommendation