Summary of Dropedge Not Foolproof: Effective Augmentation Method For Signed Graph Neural Networks, by Zeyu Zhang et al.
DropEdge not Foolproof: Effective Augmentation Method for Signed Graph Neural Networks
by Zeyu Zhang, Lu Li, Shuyan Wan, Sijie Wang, Zhiyi Wang, Zhiyuan Lu, Dong Hao, Wanli Li
First submitted to arxiv on: 29 Sep 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 The paper explores signed graphs, a novel representation of relationships between entities with marked edges indicating friendly or antagonistic interactions. The authors focus on link sign prediction, which is challenging due to graph sparsity and unbalanced triangles in Signed Graph Neural Networks (SGNNs). To address these issues, the authors propose using data augmentation (DA) techniques, specifically introducing the Signed Graph Augmentation (SGA) framework. SGA includes a structure augmentation module for identifying candidate edges and a strategy for selecting beneficial candidates, which improves SGNN training. Experimental results demonstrate that SGA boosts the performance of SGNN models by 32.3% in F1-micro on the Slashdot dataset. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about a new way to understand relationships between people or things using special graphs. These graphs have edges with signs, which can be positive (friends) or negative (enemies). The authors want to make it easier for computers to predict these relationships, but it’s hard because some graphs are very sparse (not many connections) and others have unbalanced triangles (unequal numbers of friends and enemies). They came up with a new approach called Signed Graph Augmentation, which helps improve the performance of computer models in predicting relationships. The results show that this new approach can make predictions more accurate by 32%. |
Keywords
» Artificial intelligence » Data augmentation
