Summary of Single-view Graph Contrastive Learning with Soft Neighborhood Awareness, by Qingqiang Sun et al.
Single-View Graph Contrastive Learning with Soft Neighborhood Awareness
by Qingqiang Sun, Chaoqi Chen, Ziyue Qiao, Xubin Zheng, Kai Wang
First submitted to arxiv on: 12 Dec 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Medium Difficulty Summary: Most graph contrastive learning (GCL) methods rely on cross-view contrasts, which pose challenges such as designing effective augmentations and information loss. To address this, we propose SIGNA, a single-view GCL framework that leverages soft neighborhood awareness for GCL. By using dropout to obtain structurally-related yet randomly-noised embedding pairs, we create probabilistic neighborhood contrastive learning effect. Our method also employs a normalized Jensen-Shannon divergence estimator for better contrastive learning effects. Experiments on diverse node-level tasks demonstrate that SIGNA consistently outperforms existing methods by up to 21.74% (PPI), with significant speedups in transductive learning tasks using MLPs instead of GCNs. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Low Difficulty Summary: This paper talks about a new way to learn from graph data, called Graph Contrastive Learning (GCL). Most GCL methods use two views of the same graph and compare them. But this can be tricky because it’s hard to decide what to change in each view. Our new method, SIGNA, only uses one view of the graph and makes sure that nearby nodes are connected in a way that makes sense. We tested our method on many different tasks and found that it works better than other methods by up to 21.7%. Plus, it’s faster because we can use simpler models instead of complicated ones. |
Keywords
» Artificial intelligence » Dropout » Embedding
