In order to run this project do the following

python train_baseline.py runs base regressor, without modification

python train_no_cqr.py runs basic arl model without CQR

python train.py runs the whole model

in order to change the configuration, one needs to modify the following lines in each file:

centroids = torch.tensor([(1,1),(1,1),(1,1)]).float() cov = torch.tensor([[[2.0, 1], [1, 2.0]],[[2.0, 1], [1, 2.0]],[[100,10],[10,30]]]) n_samples = [10000,10000,200]

changing the centroids and the sample ratio here changes the test.

install environment.yaml to run the repository.

citations

@article{lahoti2020fairness, title={Fairness without demographics through adversarially reweighted learning}, author={Lahoti, Preethi and Beutel, Alex and Chen, Jilin and Lee, Kang and Prost, Flavien and Thain, Nithum and Wang, Xuezhi and Chi, Ed}, journal={Advances in neural information processing systems}, volume={33}, pages={728--740}, year={2020} }

@article{romano2019conformalized, title={Conformalized quantile regression}, author={Romano, Yaniv and Patterson, Evan and Candes, Emmanuel}, journal={Advances in neural information processing systems}, volume={32}, year={2019} }