Conformal Methods for Quantifying Uncertainty in Spatiotemporal Data: A Survey

TL;DR

This survey reviews conformal methods for uncertainty quantification in deep learning, emphasizing their theoretical guarantees, techniques for improved calibration in spatiotemporal data, and importance for safe decision-making in high-risk domains.

Contribution

It provides a comprehensive overview of conformal prediction techniques, highlighting recent advances and their applications to spatiotemporal data for reliable uncertainty estimation.

Findings

Conformal methods offer strong theoretical guarantees for uncertainty quantification.

Techniques to enhance calibration and efficiency in spatiotemporal data contexts.

UQ is crucial for safe decision-making in high-stakes applications.

Abstract

Machine learning methods are increasingly widely used in high-risk settings such as healthcare, transportation, and finance. In these settings, it is important that a model produces calibrated uncertainty to reflect its own confidence and avoid failures. In this paper we survey recent works on uncertainty quantification (UQ) for deep learning, in particular distribution-free Conformal Prediction method for its mathematical properties and wide applicability. We will cover the theoretical guarantees of conformal methods, introduce techniques that improve calibration and efficiency for UQ in the context of spatiotemporal data, and discuss the role of UQ in the context of safe decision making.