**LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation** **Authors:** Xiangnan He, Kuan Deng, Xiang Wang, Yan Li, Yongdong Zhang, and Meng Wang **Abstract:** Graph Convolution Network (GCN) has become a state-of-the-art approach for collaborative filtering. However, the effectiveness of GCN for recommendation is not well understood. Existing work that adapts GCN to recommendation often lacks thorough ablation studies, which show that two common designs in GCN—feature transformation and nonlinear activation—contribute little to the performance of collaborative filtering and even degrade it. 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, which includes 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. This 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 the same experimental setting. Further analyses are provided to justify the rationality of LightGCN from both analytical and empirical perspectives. **Contributions:** - Empirically show that feature transformation and nonlinear activation have no positive effect on the effectiveness of collaborative filtering. - Propose LightGCN, which simplifies the model design by including only the most essential components in GCN for recommendation. - Empirically compare LightGCN with NGCF and demonstrate substantial improvements. **Keywords:** - Collaborative Filtering - Recommendation - Embedding Propagation - Graph Neural Network **Related Work:** - Collaborative Filtering (CF) is a prevalent technique in modern recommender systems. - Graph methods for recommendation, such as Graph Convolution Networks (GCNs), have shown promise in modeling graph structure and high-hop neighbors to guide embedding learning. **Conclusion:** LightGCN simplifies the design of GCN for collaborative filtering by removing unnecessary operations and retaining only the essential components. Experiments demonstrate that LightGCN is simpler to train, has better generalization ability, and is more effective. The insights from LightGCN can inspire future developments in recommender models, particularly in exploiting graph-based models and personalizing layer combination weights.**LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation** **Authors:** Xiangnan He, Kuan Deng, Xiang Wang, Yan Li, Yongdong Zhang, and Meng Wang **Abstract:** Graph Convolution Network (GCN) has become a state-of-the-art approach for collaborative filtering. However, the effectiveness of GCN for recommendation is not well understood. Existing work that adapts GCN to recommendation often lacks thorough ablation studies, which show that two common designs in GCN—feature transformation and nonlinear activation—contribute little to the performance of collaborative filtering and even degrade it. 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, which includes 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. This 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 the same experimental setting. Further analyses are provided to justify the rationality of LightGCN from both analytical and empirical perspectives. **Contributions:** - Empirically show that feature transformation and nonlinear activation have no positive effect on the effectiveness of collaborative filtering. - Propose LightGCN, which simplifies the model design by including only the most essential components in GCN for recommendation. - Empirically compare LightGCN with NGCF and demonstrate substantial improvements. **Keywords:** - Collaborative Filtering - Recommendation - Embedding Propagation - Graph Neural Network **Related Work:** - Collaborative Filtering (CF) is a prevalent technique in modern recommender systems. - Graph methods for recommendation, such as Graph Convolution Networks (GCNs), have shown promise in modeling graph structure and high-hop neighbors to guide embedding learning. **Conclusion:** LightGCN simplifies the design of GCN for collaborative filtering by removing unnecessary operations and retaining only the essential components. Experiments demonstrate that LightGCN is simpler to train, has better generalization ability, and is more effective. The insights from LightGCN can inspire future developments in recommender models, particularly in exploiting graph-based models and personalizing layer combination weights.