Henrik Boström: Conformal Regressors and Predictive Systems - a Gentle Introduction

Abstract

Conformal prediction is a framework that allows for controlling the error rate of any predictive model. This is achieved by turning point predictions into set predictions, which are guaranteed to include the true target with a user-controlled probability. We will in this talk focus on applying the framework to regression problems, to form so-called conformal regressors, and also on a recent extension, called conformal predictive systems, which output well-calibrated predictive distributions. For references as well as an introduction to the Python package crepes (conformal regressors and predictive systems), please visit: https://github.com/henrikbostrom/crepes