Guaranteed Coverage Prediction Intervals with Gaussian Process Regression

TL;DR

This paper introduces a conformal prediction extension for Gaussian Process Regression that guarantees accurate coverage of prediction intervals even under model misspecification, improving reliability in uncertainty estimation.

Contribution

It combines Gaussian Process Regression with Conformal Prediction to ensure valid coverage guarantees regardless of model correctness, a novel integration for reliable uncertainty quantification.

Findings

The method guarantees coverage even with model misspecification.

Experimental results show improved coverage accuracy over existing methods.

The approach maintains the benefits of GPR while providing valid prediction intervals.

Abstract

Gaussian Process Regression (GPR) is a popular regression method, which unlike most Machine Learning techniques, provides estimates of uncertainty for its predictions. These uncertainty estimates however, are based on the assumption that the model is well-specified, an assumption that is violated in most practical applications, since the required knowledge is rarely available. As a result, the produced uncertainty estimates can become very misleading; for example the prediction intervals (PIs) produced for the 95% confidence level may cover much less than 95% of the true labels. To address this issue, this paper introduces an extension of GPR based on a Machine Learning framework called, Conformal Prediction (CP). This extension guarantees the production of PIs with the required coverage even when the model is completely misspecified. The proposed approach combines the advantages of GPR with the valid coverage guarantee of CP, while the performed experimental results demonstrate its superiority over existing methods.