This repository is an implementation of the paper Split Localized Conformal Prediction.
To set up the environment, run:
conda env create -f conformal.yml
then activate the environment:
conda activate conformal
To run all experiments mentioned in the paper, first change UtilsParams.experiment in config.py into corresponding name of the experiment (available choice: "prediction", "cov_shift", "toy_plot", "nn_capacity"), then run
python -W ignore run_all_experiments.py
The results will be saved into a CSV file.