Fair Canonical Correlation Analysis
Authors: Zhuoping Zhou, Davoud Ataee Tarzanagh, Bojian Hou, Boning Tong, Jia Xu, Yanbo Feng, Qi Long, Li Shen
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental evaluation on both synthetic and real-world datasets demonstrates the efficacy of our method in reducing correlation disparity error without compromising CCA accuracy.In this section, we provide empirical results showcasing the efficacy of the proposed algorithms. |
| Researcher Affiliation | Academia | University of Pennsylvania {zhuopinz@sas.,tarzanaq@,boningt@seas.,jiaxu7@,yanbof@seas.}upenn.edu {bojian.hou,qlong,li.shen}@pennmedicine.upenn.edu |
| Pseudocode | Yes | Algorithm 1: A Multi-Objective Gradient Method for F-CCA (MF-CCA)Algorithm 2: A Single-Objective Gradient Method for F-CCA (SF-CCA) |
| Open Source Code | Yes | 2Code is available at https://github.com/Penn Shen Lab/Fair_CCA. |
| Open Datasets | Yes | National Health and Nutrition Examination Survey (NHANES). We utilized the 2005-2006 subset of the NHANES database https://www.cdc.gov/nchs/nhanesMental Health and Academic Performance Survey (MHAAPS). This dataset is available at https://github.com/marks/convert_to_csv/tree/master/sample_data.Alzheimer s Disease Neuroimaging Initiative (ADNI). We utilized AV45 (amyloid) and AV1451 (tau) positron emission tomography (PET) data from the ADNI database (http://adni.loni.usc. edu) [73, 74]. |
| Dataset Splits | No | The paper mentions training on datasets and searching for hyperparameters on a grid, but it does not explicitly specify the training, validation, or test dataset splits (e.g., percentages or sample counts). |
| Hardware Specification | Yes | Experiments are run on Intel(R) Xeon(R) CPU E5-2660. |
| Software Dependencies | No | The paper mentions using 'canoncorr function from MATLAB' and 'fminimax function from the MATLAB optimization toolbox' and 'fmincon function from the MATLAB optimization toolbox' but does not provide specific version numbers for MATLAB or its toolboxes. |
| Experiment Setup | Yes | For MF-CCA and SFCCA, the learning rate is searched on a grid in {1e 1, 5e 2, 1e 2, . . . , 1e 5}, and for SF-CCA, λ is searched on a grid in {1e 2, 1e 1, 0.5, 1, 2, . . . , 10}.Termination of algorithms occurs when the descent direction norm is below 1e 4.Synthetic Data: λ = 10, learning rate of 2e-2 in SF-CCA, and 4e-1 in MF-CCA. |