Summary of Disentangling, Amplifying, and Debiasing: Learning Disentangled Representations For Fair Graph Neural Networks, by Yeon-chang Lee et al.
Disentangling, Amplifying, and Debiasing: Learning Disentangled Representations for Fair Graph Neural Networks
by Yeon-Chang Lee, Hojung Shin, Sang-Wook Kim
First submitted to arxiv on: 23 Aug 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Social and Information Networks (cs.SI)
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 This paper proposes a novel Graph Neural Network (GNN) framework, DAB-GNN, to address fairness issues in graph representation learning. GNNs are widely used in social media and healthcare, but they often suffer from biases in node attributes and graph structure, leading to unfair predictions. The proposed framework employs disentanglement and amplification modules to isolate and minimize each type of bias, followed by a debiasing module that minimizes the distance between subgroup distributions. Experimental results on five datasets demonstrate that DAB-GNN outperforms ten state-of-the-art competitors in achieving an optimal balance between accuracy and fairness. The codebase is available at this GitHub URL. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper introduces a new way to make Graph Neural Networks (GNNs) fairer. GNNs are used to analyze data from social media, healthcare, and other areas, but they often make biased decisions because of the data’s natural biases. The proposed method, DAB-GNN, helps remove these biases by separating them out and then minimizing their effects. Tests on five different datasets show that this approach works better than ten existing methods in balancing accuracy and fairness. You can find the code for this approach at a specific GitHub link. |
Keywords
* Artificial intelligence * Gnn * Graph neural network * Representation learning
