STACI: Spatio-Temporal Aleatoric Conformal Inference

TL;DR

STACI introduces a scalable, GPU-accelerated framework combining neural networks and conformal inference to provide valid uncertainty quantification for complex spatio-temporal processes, outperforming existing methods.

Contribution

It presents a novel variational Bayesian neural network approach with a new conformal inference algorithm for non-stationary spatio-temporal GPs, enhancing scalability and accuracy.

Findings

Outperforms existing GPs and deep models in accuracy

Easily scales to datasets with millions of observations

Provides statistically valid prediction intervals

Abstract

Fitting Gaussian Processes (GPs) provides interpretable aleatoric uncertainty quantification for estimation of spatio-temporal fields. Spatio-temporal deep learning models, while scalable, typically assume a simplistic independent covariance matrix for the response, failing to capture the underlying correlation structure. However, spatio-temporal GPs suffer from issues of scalability and various forms of approximation bias resulting from restrictive assumptions of the covariance kernel function. We propose STACI, a novel framework consisting of a variational Bayesian neural network approximation of non-stationary spatio-temporal GP along with a novel spatio-temporal conformal inference algorithm. STACI is highly scalable, taking advantage of GPU training capabilities for neural network models, and provides statistically valid prediction intervals for uncertainty quantification. STACI outperforms competing GPs and deep methods in accurately approximating spatio-temporal processes and we show it easily scales to datasets with millions of observations.