Abstract:The tremendous positive driving effect of Graph Convolutional Network (GCN) and Graph Contrastive Learning (GCL) for recommender systems has become a consensus. GCN encoders are extensively used in recommendation models for capturing high-order connectivities between users and items, whereas GCL accelerates the training of recommendation tasks by adding extra supervision signals from contrastive objectives. However, little attention has been paid on corresponding theories that are truly tailored to recommendation tasks. From the technical perspective, Collaborative Filtering (CF) is seen as an important factor in recommender systems. It is applied to measure user-user, item-item, and user-item similarities rather than to achieve better clustering or node classification results. Besides, heuristic-based data augmentation may not be hold true in the field of recommender systems as it requires additional training costs and introduces noises that will corrupt the interaction graph structure and the semantic information of nodes. To tackle these limitations, we propose a novel Embedding-less Graph Collaborative Filtering (EGCF) for recommendation, which is tailor-made for the problem mentioned for CF and further simplifies existing solutions. Structurally, it consists of two parts: embedding-less GCN and embedding-less GCL. The former improves user-item affinity by streamlining user-type embeddings and carrying out iterative graph convolution. And the latter utilizes three-type contrastive objectives to directly measure the alignment and the uniformity of users, items, and interaction pairs, respectively, avoiding any type of data augmentation or multi-view construction. Even though EGCF has been extremely streamlined, extensive experimental results on three classical datasets demonstrate the effectiveness of EGCF in terms of recommendation accuracy and training efficiency. The code and used datasets are released at https://github.com/BlueGhostYi/ID-GRec.

Dynamic Evolution of Multi-Graph Based Collaborative Filtering for Recommendation Systems.

Improved Metric-Based Recommender by Historical Interactions.

Dynamic Graph Collaborative Filtering

Social Recommendation With Evolutionary Opinion Dynamics

Dyn-GCN: Graph Embedding Via Dynamic Evolution and Graph Convolutional Network for Personal Recommendation

Graph Heterogeneous Multi-Relational Recommendation

TCGC: Temporal Collaboration-Aware Graph Co-Evolution Learning for Dynamic Recommendation

Learning Shared Representations for Recommendation with Dynamic Heterogeneous Graph Convolutional Networks

Exploiting Structural and Temporal Influence for Dynamic Social-Aware Recommendation

Unified Collaborative Filtering over Graph Embeddings

Graph Representation-Based Deep Multi-View Semantic Similarity Learning Model for Recommendation

Dynamic Recommendation Based on Graph Diffusion and Ebbinghaus Curve

MGRec: Multi-Graph Fusion for Recommendation.

Simplify to the Limit! Embedding-less Graph Collaborative Filtering for Recommender Systems

Graph Collaborative Filtering Based on Dual-Message Propagation Mechanism.

Multi-Behavior Dynamic Contrastive Learning for Recommendation

Graph Collaborative Signals Denoising and Augmentation for Recommendation

Collaborative Filtering Recommendation Method Based on Graph Convolutional Neural Networks

SocialGCN: an Efficient Graph Convolutional Network Based Model for Social Recommendation

Dual Channel Hypergraph Collaborative Filtering

A Personalized Collaborative Filtering Recommendation System Based on Bi-Graph Embedding and Causal Reasoning