# DAVIDSTUTZ

## Conformal Training

### Abstract

Modern deep learning based classifiers show very high accuracy on test data but this does not provide sufficient guarantees for safe deployment, especially in high-stake AI applications such as medical diagnosis. Usually, predictions are obtained without a reliable uncertainty estimate or a formal guarantee. Conformal prediction (CP) addresses these issues by using the classifier's probability estimates to predict confidence sets containing the true class with a user-specified probability. However, using CP as a separate processing step after training prevents the underlying model from adapting to the prediction of confidence sets. Thus, this paper explores strategies to differentiate through CP during training with the goal of training model with the conformal wrapper end-to-end. In our approach, conformal training (ConfTr), we specifically "simulate" conformalization on mini-batches during training. We show that CT outperforms state-of-the-art CP methods for classification by reducing the average confidence set size (inefficiency). Moreover, it allows to "shape" the confidence sets predicted at test time, which is difficult for standard CP. On experiments with several datasets, we show ConfTr can influence how inefficiency is distributed across classes, or guide the composition of confidence sets in terms of the included classes, while retaining the guarantees offered by CP.

### Paper

The paper is available on ArXiv:

@inproceedings{Stutz2022ICLR,
author    = {David Stutz and Krishnamurthy and Dvijotham and Ali Taylan Cemgil and Arnaud Doucet},
title     = {Learning Optimal Conformal Classifiers},
booktitle = ICLR,
year      = {2022},
}


### Poster

The below poster is also available as PDF:

### Talks

This work was presented at the Seminar on Distribution-Free Statistics organized by Anastasios Angelopoulos:

### Code

Jax-based code for implementing conformal training in included in appendix O:

### News & Updates

Apr 6, 2022. The poster if now available.

Mar 7, 2022. The camera ready version is now on ArXiv.

Jan 20, 2022. Our paper was accepted at ICLR 2022!

Oct 18, 2021. The paper is available as pre-print on ArXiv.