This repository contains the code used for the numerical experiments of the following paper:

U. Gazin, R. Heller, E. Roquain, A. Solari "Powerful batch conformal prediction for classification".

The code is under MIT Licence but please refer to and cite the above paper if you use it for academic purposes.

The folder "Source" contains two python package (black_boxes.py and nn.py), which contains some general codes for machine learning algorithms. They can be found in A. Marandon page: https://github.com/arianemarandon/infoconf, and are based on code from the paper "Classification with Valid and Adaptive Coverage" by Romano, Sesia and Candès (Neurips 2020).

This folder contains some general code to use the batch conformal prediction methodology, with the various procedure decribed in the paper, some applications on the CIFAR and USPS dataset, and the code to obtain the illustrations on real data sets (Section 4.2). The notebook "Python_Batch_GHRS25.ipynb" the experiments on the USPS and CIFAR datset. The python file "GeneralBatchCode.py" contains the general methodology and the file "LargeSampleCode.py" contains the method to compare the different methods with Monte-Carlo simulations. Inside the folder "Illustration", the notebook "Illustration_AISTAS25.ipynb" contains the code to make the illustration in the paper "Powerful batch conformal prediction for classification". The ".npy" are the result of some experiments donne with the codes from "Python_Batch_GHRS25.ipynb".

This folder contains the code for the Bivariate Normal Simulations (Section 4.1) simulations, to reproduce the tables in F.4 ("NumericalExperimentsSectionGaussianMultivariateSetting") and in D.4 ("NumericalExperimentsSectionLargeBatches") in R Markdown format.