Spatiotemporal System Forecasting with Irregular Time Steps via Masked Autoencoder

arXiv cs.LG / 3/27/2026

📰 NewsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper addresses forecasting high-dimensional spatiotemporal dynamical systems when observations occur at irregular time steps, which commonly degrade the accuracy of existing data-driven models.
It proposes a Physics-Spatiotemporal Masked Autoencoder that combines convolutional autoencoders for spatial representation with a masked autoencoder optimized for irregular time series using attention to reconstruct the full physical sequence in one pass.
The method is designed to avoid explicit data imputation while maintaining the physical coherence of the generated spatiotemporal fields.
Experiments on simulated datasets and real ocean temperature data show improved prediction accuracy, robustness to nonlinear dynamics, and better computational efficiency than conventional convolutional and recurrent approaches.
The authors position the approach as applicable to domains such as climate modeling, fluid dynamics, ocean forecasting, environmental monitoring, and scientific computing without requiring heavy domain-specific knowledge.

Abstract

Predicting high-dimensional dynamical systems with irregular time steps presents significant challenges for current data-driven algorithms. These irregularities arise from missing data, sparse observations, or adaptive computational techniques, reducing prediction accuracy. To address these limitations, we propose a novel method: a Physics-Spatiotemporal Masked Autoencoder. This method integrates convolutional autoencoders for spatial feature extraction with masked autoencoders optimised for irregular time series, leveraging attention mechanisms to reconstruct the entire physical sequence in a single prediction pass. The model avoids the need for data imputation while preserving physical integrity of the system. Here, 'physics' refers to high-dimensional fields generated by underlying dynamical systems, rather than the enforcement of explicit physical constraints or PDE residuals. We evaluate this approach on multiple simulated datasets and real-world ocean temperature data. The results demonstrate that our method achieves significant improvements in prediction accuracy, robustness to nonlinearities, and computational efficiency over traditional convolutional and recurrent network methods. The model shows potential for capturing complex spatiotemporal patterns without requiring domain-specific knowledge, with applications in climate modelling, fluid dynamics, ocean forecasting, environmental monitoring, and scientific computing.