Fairness via Group Contribution Matching
Authors: Tianlin Li, Zhiming Li, Anran Li, Mengnan Du, Aishan Liu, Qing Guo, Guozhu Meng, Yang Liu
IJCAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experiments show that our GCM effectively improves fairness and outperforms other methods significantly. Extensive experiments on three public datasets show that our GCM method could effectively improve fairness and outperform other baseline methods significantly. In our experiments, we use two tabular benchmarks (Adult and COMPAS) and one image dataset (Celeb A) that are all for binary classification tasks |
| Researcher Affiliation | Academia | Tianlin Li1 , Zhiming Li1 , Anran Li1, , Mengnan Du2 , Aishan Liu3 , Qing Guo4,5 , Guozhu Meng 6 and Yang Liu 7,1, 1Nanyang Technological University, Singapore, 2New Jersey Institute of Technology, USA, 3Beihang University, China, 4Institute of High Performance Computing (IHPC), A*STAR, Singapore, 5Centre for Frontier AI Research (CFAR), A*STAR, Singapore, 6SKLOIS, Institute of Information Engineering, Chinese Academy of Sciences, China, 7Zhejiang Sci-Tech University, China |
| Pseudocode | No | The paper describes methods with mathematical equations and text, but no explicitly labeled 'Algorithm' or 'Pseudocode' block. |
| Open Source Code | No | The paper does not contain any statement about releasing source code or a link to a repository for their method. |
| Open Datasets | Yes | In our experiments, we use two tabular benchmarks (Adult and COMPAS) and one image dataset (Celeb A) that are all for binary classification tasks: ❶Adult [Dua and Graff, 2017]. ❷Celeb A [Liu et al., 2015]. ❸COMPAS [Mele and many others, 2017 2021]. ... We further extend experiments on two datasets, Colored MNIST [Arjovsky et al., 2019], and CIFAR-10S [Wang et al., 2020] to show the performance of GCM. |
| Dataset Splits | No | The paper mentions using training and test sets but does not specify validation splits or other detailed splitting methodologies for reproducibility. It says 'For the adult dataset, we follow the settings in [Chuang and Mroueh, 2021] for data preprocessing.' without detailing the splits within this paper. |
| Hardware Specification | No | No specific hardware (GPU/CPU models, memory) is mentioned. |
| Software Dependencies | No | The paper mentions using Adam as the optimizer and MLP, Alex Net, Res Net-18 as models, but does not specify software versions for any libraries (e.g., PyTorch, TensorFlow, scikit-learn) or Python version. |
| Experiment Setup | Yes | We use Adam as the learning optimizer and the batch size is set as 2000 following the setting in [Chuang and Mroueh, 2021]. The learning rate is set as 0.001. For the Celeb A dataset, We use Adam as the learning optimizer and the batch size is set as 128. The learning rate is set as 0.0001. For the COMPAS dataset, we use Adam as the learning optimizer, and the batch size is set as 2000. The learning rate is set as 0.001. |