Abstract:Graph Neural Networks (GNNs) have become powerful tools in modeling graph-structured data in recommender systems. However, real-life recommendation scenarios usually involve heterogeneous relationships (e.g., social-aware user influence, knowledge-aware item dependency) which contains fruitful information to enhance the user preference learning. In this paper, we study the problem of heterogeneous graph-enhanced relational learning for recommendation. Recently, contrastive self-supervised learning has become successful in recommendation. In light of this, we propose a Heterogeneous Graph Contrastive Learning (HGCL), which is able to incorporate heterogeneous relational semantics into the user-item interaction modeling with contrastive learning-enhanced knowledge transfer across different views. However, the influence of heterogeneous side information on interactions may vary by users and items. To move this idea forward, we enhance our heterogeneous graph contrastive learning with meta networks to allow the personalized knowledge transformer with adaptive contrastive augmentation. The experimental results on three real-world datasets demonstrate the superiority of HGCL over state-of-the-art recommendation methods. Through ablation study, key components in HGCL method are validated to benefit the recommendation performance improvement. The source code of the model implementation is available at the link <a class="link-external link-https" href="https://github.com/HKUDS/HGCL" rel="external noopener nofollow">this https URL</a>.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper primarily explores how to enhance relationship learning in recommendation systems using Heterogeneous Graph Contrastive Learning (HGCL) to improve recommendation performance. Specifically, the paper attempts to address the following issues: 1. **Integrating Heterogeneous Information**: - Many existing Graph Neural Network (GNN)-based recommendation systems focus only on the homogeneous interaction between users and items, neglecting heterogeneous relationships such as social network connections. The paper proposes a method to enhance recommendation systems by integrating heterogeneous relationship information. 2. **Data Sparsity Issue**: - Existing heterogeneous graph neural networks typically require a large number of training labels to generate high-quality user/item embeddings. However, in practical applications, label data is usually sparse. To address this, the paper introduces Contrastive Self-Supervised Learning to mitigate the data sparsity issue. 3. **Personalized Knowledge Transfer**: - The impact of heterogeneous information on users and items varies due to individual differences. For example, a user's social influence may vary based on their personal characteristics. Therefore, the paper proposes a Meta Network to achieve personalized knowledge transfer, allowing each user and item to receive specific information enhancement. ### Main Contributions 1. **Heterogeneous Graph Contrastive Learning Framework**: - A new framework named HGCL is proposed to integrate heterogeneous information in recommendation systems and optimize it under the graph contrastive learning paradigm. 2. **Adaptive Contrastive Enhancement**: - By combining contrastive learning with adaptive enhancement through a Meta Network, the framework achieves user- and item-specific knowledge transfer, enhancing cross-view information transfer capabilities. 3. **Experimental Validation**: - Extensive experiments were conducted on three real-world datasets, validating the significant improvement in recommendation performance brought by HGCL. In summary, this paper aims to significantly improve the performance of recommendation systems by integrating heterogeneous information and optimizing it under a contrastive learning framework.

Heterogeneous Graph Contrastive Learning for Recommendation

Semantic-enhanced Heterogeneous Graph Contrastive Learning for Recommendation

Heterogeneous Graph Neural Networks using Self-supervised Reciprocally Contrastive Learning

Graph Contrastive Learning with Kernel Dependence Maximization for Social Recommendation

LightGCL: Simple Yet Effective Graph Contrastive Learning for Recommendation

GNNCL: A Graph Neural Network Recommendation Model Based on Contrastive Learning

Heterogeneous Graph Contrastive Learning with Dual Aggregation Scheme and Adaptive Augmentation

Adaptive Graph Contrastive Learning for Recommendation

Intent-guided Heterogeneous Graph Contrastive Learning for Recommendation

Globally Informed Graph Contrastive Learning for Recommendation

Heterogeneous Adaptive Preference Learning for Recommendation

Graph Learning Augmented Heterogeneous Graph Neural Network for Social Recommendation

Generative Contrastive Graph Learning for Recommendation

SSGCL: Simple Social Recommendation with Graph Contrastive Learning

Generative-Contrastive Graph Learning for Recommendation

Session-based Recommendation Via Contrastive Learning on Heterogeneous Graph

Graph Heterogeneous Multi-Relational Recommendation

Enhancing Graph Contrastive Learning with Reliable and Informative Augmentation for Recommendation

Contrastive Graph Learning for Social Recommendation

Heterogeneous Graph Contrastive Learning With Augmentation Graph

Hypergraph Enhanced Contrastive Learning for News Recommendation.