Spatio-temporal probabilistic forecast using MMAF-guided learning

TL;DR

This paper introduces MMAF-guided learning, a Bayesian approach for spatio-temporal forecasting using neural networks that incorporate causal structure, demonstrating effective calibration and competitive performance with shallow architectures.

Contribution

The paper develops a theory-guided Bayesian methodology for spatio-temporal neural networks that enforces causal structure, enabling calibrated forecasts with simpler models.

Findings

Forecasts remain calibrated across multiple horizons.

Shallow neural networks can outperform deep architectures in this setting.

Method is validated on synthetic and real data.

Abstract

We present a theory-guided generalized Bayesian methodology for spatio-temporal raster data, which we use to train an ensemble of stochastic feed-forward neural networks with Gaussian-distributed weights. The methodology incorporates the dependence and causal structure of a spatio-temporal Ornstein-Uhlenbeck process into training and inference by enforcing constraints on the design of the data embedding and the related optimization routine. In inference mode, the networks are employed to generate causal ensemble forecasts by applying different initial conditions at different horizons. We call this workflow MMAF-guided learning. Experiments conducted on both synthetic and real data demonstrate that our forecasts remain calibrated across multiple time horizons. Moreover, we show that on such data, shallow feed-forward architectures can achieve performance comparable to, and in some cases better than, convolutional or diffusion deep learning architectures used in probabilistic forecasting tasks.