· 2020
LightGCN
- 4,168citations
- 2020year
Short summary
Researchers found that common neural network operations in Graph Convolutional Networks (GCNs), specifically feature transformation and nonlinear activation, are not only unnecessary but detrimental to recommendation performance.
AI-generated from the title and abstract; the full text is not read.
Key points
- Feature transformation and nonlinear activation in GCNs offer minimal benefit for recommendation tasks.
- Including these neural network operations can degrade recommendation performance.
- A simplified GCN architecture (LightGCN) is proposed, implying better performance and training efficiency.
AI-generated from the title and abstract; the full text is not read.
Abstract
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.
The authors' abstract, as published at the source., 2020 · DOI ↗
Continue with a free account
Ask the paper: 3 free questions a day about this paper; save it, get its citation, new summaries every day for your field. Takeaways are Premium.
Continue free on the webSign in with Google or Apple; no card needed. You come back to this paper.
On your phone:
Field: Information Systems
Information SystemsComputer Science