Understanding and Mitigating Accuracy Disparity in Regression
Authors: Jianfeng Chi, Yuan Tian, Geoffrey J. Gordon, Han Zhao
ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | To corroborate the effectiveness of our proposed algorithms in reducing accuracy disparity, we conduct experiments on five benchmark datasets. Experimental results suggest that our proposed algorithms help to mitigate accuracy disparity while maintaining the predictive power of the regression models. |
| Researcher Affiliation | Academia | 1Department of Computer Science, University of Virginia 2Machine Learning Department, Carnegie Mellon University 3Department of Computer Science, University of Illinois at Urbana-Champaign. |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Our code is publicly available at: https://github.com/JFChi/Understanding-and -Mitigating-Accuracy-Disparity-in-Regressi |
| Open Datasets | Yes | Datasets We conduct experiments on five benchmark datasets: the Adult dataset (Dua and Graff, 2017), COMPAS dataset (Dieterich et al., 2016), Communities and Crime dataset (Dua and Graff, 2017), Law School dataset (Wightman and Ramsey, 1998) and Medical Insurance Cost dataset (Lantz, 2013). |
| Dataset Splits | No | The paper mentions using benchmark datasets and averaging results over ten random seeds but does not specify the exact training, validation, and test splits (e.g., percentages or sample counts). |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., CPU/GPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions implementing algorithms using Pytorch but does not provide specific version numbers for PyTorch or other software dependencies. |
| Experiment Setup | Yes | For each dataset, we perform controlled experiments by fixing the regression model architectures to be the same. We train the regression models via minimizing mean squared loss. Among all methods, we vary the trade-off parameter (i.e., λ in CENET and WASSERSTEINNET and in BGL and COD) and report and the corresponding R2 scores and the error gap values. We refer readers to Appendix B for detailed hyper-parameter settings in our experiments and Appendix C for additional experimental results. |