LightGCN
LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation
Xiangnan He, Kuan Deng, Xiang Wang, Yan Li, Yongdong Zhang and Meng Wang · SIGIR 2020
Propagate collaborative signals through a symmetrically normalized user-item graph and aggregate multiple layers including initial embeddings.
What you will implement
Dense bipartite graphs, linear propagation, uniform layer aggregation and 160 actual synthetic training updates; no large sparse training or paper dataset benchmark.
- 69. LightGCN: Normalized Bipartite Graphs and Collaborative Propagation
Propagate collaborative signals through a user-item graph and differentiate the aggregation including the initial layer.
Symmetric normalization of a dense bipartite graph, linear multi-hop propagation and uniform layer averaging; BPR loss and tiny training loop are provided.