Stable Fair Graph Representation Learning with Lipschitz Constraint
Authors: Qiang Chen, Zhongze Wu, Xiu Su, Xi Lin, Zhe Qu, Shan You, Shuo Yang, Chang Xu
ICML 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments conducted on three real-world datasets demonstrate that SFG is stable and outperforms other state-of-the-art adversarial-based methods in terms of both fairness and utility performance. |
| Researcher Affiliation | Collaboration | 1Central South University, Changsha, Hunan, China 2Shanghai Jiaotong University, Shanghai, China 3Sense Time Research, Shanghai, China 4Harbin Institute of Technology (Shenzhen), Shenzhen, China 5University of Sydney, Sydney, Australia. |
| Pseudocode | Yes | Algorithm 1 The accelerated DFB algorithm with mask generator |
| Open Source Code | Yes | Codes are available at https://github.com/sh-qiangchen/SFG. |
| Open Datasets | Yes | We conduct experiments on three commonly used datasets(Dong et al., 2022), including German, Bail, and Credit. |
| Dataset Splits | No | The paper mentions tuning hyperparameters on a validation set and running experiments multiple times, but does not specify the exact percentages or methodology used for splitting the datasets into training, validation, and test sets. For example, 'For a fair comparison, we tuned the hyperparameters for all methods according to metric = ACC + AUC + F1 DP EO on the validation set.' |
| Hardware Specification | No | The paper does not provide specific hardware details such as GPU models, CPU types, or memory specifications used for running the experiments. It only describes the experimental setup in terms of models and hyperparameters. |
| Software Dependencies | No | The paper does not specify the version numbers for any software dependencies, such as programming languages, deep learning frameworks (e.g., PyTorch, TensorFlow), or other libraries, that would be needed to replicate the experiment. |
| Experiment Setup | Yes | For SFG, we use a 2-layer Graph SAGE encoder with hidden dimensions 16 and set the number of generated fair feature views µ = 10 for all datasets. We set the range of the Lipschitz constant τ is {1, 2, 4, 5, 6, 20, 50}, which should not be larger than the unconstrained counterpart, for German, Bail, Credit. |