Summary of Disentangled Hyperbolic Representation Learning For Heterogeneous Graphs, by Qijie Bai et al.
Disentangled Hyperbolic Representation Learning for Heterogeneous Graphs
by Qijie Bai, Changli Nie, Haiwei Zhang, Zhicheng Dou, Xiaojie Yuan
First submitted to arxiv on: 14 Jun 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 The proposed Dis-H^2GCN is a novel approach to embedding complex real-world systems into low-dimensional spaces. This method addresses two challenges: mixing structural and semantic information and distributional mismatch between data and embedding spaces. To disentangle these features, the model uses mutual information minimization and discrimination maximization constraints, along with independent message propagation for each edge type. The entire model is constructed upon hyperbolic geometry to narrow the gap between data distributions and representing spaces. This approach is evaluated on five real-world heterogeneous graph datasets across two downstream tasks: node classification and link prediction. The results demonstrate the superiority of Dis-H^2GCN over state-of-the-art methods, showcasing its effectiveness in disentangling and representing heterogeneous graph data in hyperbolic spaces. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Dis-H^2GCN is a new way to represent complex systems on low-dimensional spaces. It solves two problems: mixing information from structures and semantics, and different distributions of data and representations. The method uses special math constraints to separate these features and then connects them using messages that only consider specific types of edges. The model also works in hyperbolic geometry to make the representations closer to real-world data. This approach is tested on five datasets for two tasks: classifying nodes and predicting links. The results show that Dis-H^2GCN performs better than other methods, making it a good way to represent complex systems. |
Keywords
* Artificial intelligence * Classification * Embedding * Semantics
