Abstract:In recommendation tasks, we model user preferences by learning node representations (i.e., user and item embeddings) based on the observed user-item interaction data, which is a bipartite graph. G raph N eural N etworks ( GNN s) are widely used to refine the representations by exploring the topology of the graph: embeddings of neighbors are propagated to each node to reconstruct its embeddings. However, the propagation strategy in existing GNNs is empirical and defective: (1) a substantial proportion of links are missed in the sparse observed graph, which causes ineffective and biased propagation; and (2) the propagation weights are determined by a coarse pre-defined rule, which only takes the degree of nodes into consideration. In this paper, we propose a dense and data-driven propagation mechanism for GNNs. Considering the graph we use to propagate embeddings in recommendation tasks is extremely sparse, we complement it and use the predicted graph as the new propagation tool. We learn the propagation matrix from the data, and propose a S elf-propagation G raph N eural N etwork ( SGNN ). Since it is very space- and time-consuming to maintain a large and dense propagation matrix, we factorize it for storing and updating. In this paper, we propose three methods to complete the sparse graph and construct the propagation matrix: (1) we complete the graph based on a recommendation model; (2) we measure the node distance based on spectral clustering; (3) we predict missing links of the graph based on predictive embeddings. In SGNN, the embeddings can be propagated to not only the observed neighbors, but also the potential yet unobserved neighbors, and the propagation weights are learned based on the connection strength. Comprehensive experiments on three real-world datasets demonstrate the effectiveness and efficiency of our proposed model: SGNN outperforms recent state-of-the-art GNNs significantly. Codes are available on https://github.com/Wenhui-Yu/LCFN .

XGCN: a library for large-scale graph neural network recommendations

Self-Propagation Graph Neural Network for Recommendation

Xgcn: an Extreme Graph Convolutional Network for Large-scale Social Link Prediction

Scalable and Efficient Full-Graph GNN Training for Large Graphs

IntentGC: a Scalable Graph Convolution Framework Fusing Heterogeneous Information for Recommendation

Attributed Graph Neural Networks for Recommendation Systems on Large-Scale and Sparse Graph

GNNRec: gated graph neural network for session-based social recommendation model

Friend Recommendations with Self-Rescaling Graph Neural Networks

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

Linear-Time Graph Neural Networks for Scalable Recommendations

ByteGNN: Efficient Graph Neural Network Training at Large Scale

SGNNRec: A Scalable Double-Layer Attention-Based Graph Neural Network Recommendation Model

Ada-GNN: Adapting to Local Patterns for Improving Graph Neural Networks

UltraGCN: Ultra Simplification of Graph Convolutional Networks for Recommendation

TransGNN: Harnessing the Collaborative Power of Transformers and Graph Neural Networks for Recommender Systems

SVD-GCN: A Simplified Graph Convolution Paradigm for Recommendation

LazyGNN: Large-Scale Graph Neural Networks via Lazy Propagation

Large-scale Online Multi-View Graph Neural Network and Applications

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

A Comprehensive Study on Large-Scale Graph Training: Benchmarking and Rethinking

A Scalable Social Recommendation Framework with Decoupled Graph Neural Network