Observation-Aligned Mask Priors for Learning Physical Dynamics from Authentic Occlusions

TL;DR

This paper introduces a novel framework that learns authentic occlusion mask distributions to improve physical dynamics modeling from incomplete satellite observations, outperforming heuristic methods.

Contribution

It proposes Observation-Aligned Mask Priors, combining Bayesian Flow Networks and diffusion models to generate realistic, sample-specific masks for training from incomplete data.

Findings

Improved MSE and PSNR on oceanographic datasets with authentic satellite occlusions.

The mask intersection approach prevents zero-query dead zones and enhances model robustness.

Demonstrates the effectiveness of learned mask priors over heuristic masking methods.

Abstract

Learning physical dynamics directly from incomplete observations is challenging because authentic occlusions are structured, sample-dependent, and often missing not at random, whereas existing methods typically rely on heuristic masking rules or predefined mask distributions. We propose Observation-Aligned Mask Priors, a framework that learns the distribution of authentic observation masks and uses it to construct context-query partitions for training from incomplete data. Specifically, we pretrain a Bayesian Flow Network (BFN) on binary observation masks to capture real occlusion topologies, then guide BFN sampling with a globally normalized cross-entropy objective to generate sample-specific masks aligned with each sparse observation. The intersection between the guided mask and the observed mask defines the context, and the remaining observed entries become query targets for a diffusion-based reconstruction model. We show that this intersection-based partitioning gives every valid observed dimension a strictly positive probability of being queried, preventing zero-query dead zones and local generative collapse. Experiments on three real-world oceanographic datasets with authentic satellite occlusions, across resolutions up to 256$\times$256, show consistent improvements over strong diffusion baselines in MSE and PSNR. These results demonstrate that learning mask priors from authentic occlusions is an effective alternative to heuristic masking for learning from incomplete physical observations without access to fully observed fields.