Summary of Gtc: Gnn-transformer Co-contrastive Learning For Self-supervised Heterogeneous Graph Representation, by Yundong Sun et al.
GTC: GNN-Transformer Co-contrastive Learning for Self-supervised Heterogeneous Graph Representation
by Yundong Sun, Dongjie Zhu, Yansong Wang, Zhaoshuo Tian
First submitted to arxiv on: 22 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Information Retrieval (cs.IR)
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 proposes a novel framework that combines Graph Neural Networks (GNNs) and Transformers to eliminate the over-smoothing problem in GNNs. The authors argue that while GNNs are great at local information aggregation, they struggle with modeling global information and multi-hop interactions. In contrast, Transformers can model global information via self-attention mechanisms. To integrate both, the paper proposes a collaborative learning scheme called GTC, which encodes node information from different views using both GNNs and Transformers. The Transformer branch uses Metapath-aware Hop2Token and CG-Hetphormer to attendively encode neighborhood information from different levels. Experimental results on real datasets show that GTC outperforms state-of-the-art methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper combines two powerful tools, Graph Neural Networks (GNNs) and Transformers, to solve a big problem in graph learning. Right now, GNNs are great at understanding what’s happening nearby, but they struggle with seeing the bigger picture. Transformers can do that, but they’re not as good at local information. The researchers want to see if they can combine these two strengths to create something better. They came up with a new way of learning called GTC, which uses both GNNs and Transformers to understand graphs from different angles. This helps them avoid a problem called over-smoothing, where the model gets too good at understanding local information but forgets about global relationships. The results show that this new approach works really well! |
Keywords
* Artificial intelligence * Self attention * Transformer
