EtashGuha/tail_cp

This code was run on Python version 3.10. To initialize your environment, please run

bash setup.sh

if you have a GPU. Otherwise, please run

bash setup_for_cpu.sh

Then, to recreate any of our experiments, please run

python main.py -c cfgs/[name of yaml] --seed [seed]

All of our experimental setups are saved in the cfgs folder. We detail all the folders and the experiments in each. cfgs/best: all of our method's experiments cfgs/cb_baseline: experiments for Conformal Bayes cfgs/chr: experiments for CHR cfgs/cqr: experiments for CQR cfgs/mle: experiments for MLE ablation cfgs/mle_entropy: experiments for MLE with Entropy ablation cfgs/heteroskedastic: experiments to recreate the heteroskedastic figures cfgs/lei_baseline: experiments to recreate the KDE from Lei 2014 cfgs/no_entropy: experiments for ablation without entropy cfgs/ridge_baseline: experiments for Lasso Method from Lei 2019

For details on how to get MEPs data, please refer to https://github.com/yromano/cqr/tree/master/get_meps_data