LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation

By Xiangnan He, Kuan Deng, Xiang Wang, Yan Li, Yongdong Zhang, Meng Wang. Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, 2020

5,414 citations recorded by the source corpus.

Graph Convolution Network (GCN) has become new state-of-the-art for collaborative filtering. Nevertheless, the reasons of its effectiveness for recommendation are not well understood. Existing work that adapts GCN to recommendation lacks thorough ablation analyses on GCN, which is originally designed for graph classification tasks and equipped with many neural network operations. However, we empirically find that the two most common designs in GCNs -- feature transformation and nonlinear activation -- contribute little to the performance of collaborative filtering. Even worse, including them adds to the difficulty of training and degrades recommendation performance. In this work, we aim to simplify the design of GCN to make it more concise and appropriate for recommendation. We propose a new model named LightGCN, including only the most essential component in GCN -- neighborhood aggregation -- for collaborative filtering. Specifically, LightGCN learns user and item embeddings by linearly propagating them on the user-item interaction graph, and uses the weighted sum of the embeddings learned at all layers as the final embedding. Such simple, linear, and neat model is much easier to implement and train, exhibiting substantial improvements (about 16.0% relative improvement on average) over Neural Graph Collaborative Filtering (NGCF) -- a state-of-the-art GCN-based recommender model -- under exactly the same experimental setting. Further analyses are provided towards the rationality of the simple LightGCN from both analytical and empirical perspectives.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Information Systems

View Paper: https://www.semanticscholar.org/paper/3024f58826a5bce3378af94f677e8fb90cbb49e0

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Information Systems

Citations in the corpus

Self-Supervised Learning for Recommender Systems: A Survey (2022, methodology): “Itgeneratestwoaugmentedgraphswiththesame type of augmentation operator, and learns node embeddings using a shared graph LightGCN encoder f θ [84].”
Graph Neural Networks in Recommender Systems: A Survey (2020, methodology): “LightGCN [19] arxiv 2020 √ Bipartite graph variant of GCN (w/o activation & transformation)”
Denoising Implicit Feedback for Recommendation (2020, methodology): “We followed former work [6, 16] to split the dataset into training, validation, and testing (see Table 2 for statistics).”
Bias and Debias in Recommender System: A Survey and Future Directions (2020, methodology): “CDAE [164], NCF [39], LightGCN [165]) apply the uniform negative sampler; Yu et al.”
Disentangling User Interest and Conformity for Recommendation with Causal Embedding (2021, methodology): “Experiments on both MF and GCN show that it is necessary to use embeddings rather than scalar values for conformity modeling.”
LLMRec: Large Language Models with Graph Augmentation for Recommendation (2023, methodology): “To inject high-order [32] collaborative connectivity into augmented features f A ,𝑢 and f A ,𝑖 , LLMRec employs light weight GNNs [12] as the encoder.”
Graph Neural Networks for Recommender System (2022, methodology): “The development of recommender systems generally includes three stages, shallow models [16, 18], neural network (NN) -based models [9, 11, 13], and GNN-based models [2, 7, 12, 21].”
Parameter-Efficient Transfer from Sequential Behaviors for User Modeling and Recommendation (2020, methodology): “ream tasks, we randomly split the target dataset into training (70%), validation (3%) and testing (27%) sets. We use two popular top-5 metrics — MRR@5 (Mean Reciprocal Rank) [41] and HR@5 (Hit Ratio) [12, 13] — for the cold-start recommendation datasets (i.e. ColdRecs), and the classification accura”
LightGCL: Simple Yet Effective Graph Contrastive Learning for Recommendation (2023, methodology): “• LightGCN (He et al., 2020) adopts a simpliﬁed GCN structure without embedding weight matrices and non-linear projection.”
Knowledge Graph Contrastive Learning for Recommendation (2022, methodology): “…(MF) to neural network-based techniques for latent user and item embedding projection, such as Autoencoder-based approaches ( e.g., Autorec [28]), attentive CF mechanisms ( e.g., ACF [4]), as well as recently developed CF models built upon graph convolutional architectures ( e.g., Light-GCN [11]).”
Advances and Challenges in Conversational Recommender Systems: A Survey (2021, background): “Recently, more complicated but powerful neural networks have been developed, including Wide&Deep [28], neural collaborative filtering (NCF) [58], deep interest network (DIN) [218], tree-based deep model (TDM) [227], and graph convolutional networks [199, 59].”
Mining Latent Structures for Multimedia Recommendation (2021, methodology): “These graph-based recommender systems [14, 41, 43] achieve great success and obtain state-of-the-art performance.”
Recommender Systems in the Era of Large Language Models (LLMs) (2023, neutral): “Considering users’ online behaviors ( e.g., chick, purchase, socializing) as graph-structured data, Graph Neural Networks (GNNs) have emerged as powerful representation learning techniques to learn user and item representations [1], [11].”
TALLRec: An Effective and Efficient Tuning Framework to Align Large Language Model with Recommendation (2023, methodology): “They regard the LLM as a toolformer [42] of traditional recommendation models (such as MF [27] and LightGCN [16]), i.”
Self-supervised Graph Learning for Recommendation (2020, methodology): “trainable parameters, the space complexity remains the same as LightGCN [17].”