Spatiotemporal Imputation with Graph-Informed Flow Matching

2026-06-04

AI Summary

This work addresses the challenge of missing data in spatiotemporal sequences, where existing methods are often hindered by error accumulation, computational inefficiency, or unrealistic Gaussian priors. To overcome these limitations, we propose GiFlow, a novel framework that replaces conventional Gaussian assumptions with a graph-structured prior to better capture underlying spatiotemporal relationships. GiFlow integrates spatial attention, temporal attention, and a spatiotemporal propagation mechanism within a hybrid vector field model, enabling joint modeling of complex dependencies across space and time. This design substantially simplifies trajectory generation while enhancing representational capacity. Extensive experiments on multiple synthetic and real-world datasets demonstrate that GiFlow significantly outperforms state-of-the-art methods in spatiotemporal imputation, achieving both superior accuracy and computational efficiency.

Abstract

Missing data is a common challenge in spatiotemporal systems, arising in applications such as air quality monitoring and urban traffic management. Traditional machine learning approaches, like recurrent and graph neural networks, rely on iterative propagation, which tends to accumulate errors over time and space. Recent diffusion-based methods mitigate error propagation but require iterative sampling and often depend on problem-agnostic Gaussian priors, limiting both efficiency and effectiveness. To address these limitations, we propose GiFlow, a Graph-Informed Flow Matching framework for spatiotemporal imputation. GiFlow replaces the typical Gaussian prior with a graph-informed prior constructed via spatiotemporal filtering of observable signals, which better aligns the source distribution to the target and thereby simplifies the generation trajectory. The flow field is parameterized by a hybrid vector field model that integrates spatial attention, temporal attention, and spatiotemporal propagation, enabling joint modeling of spatial and temporal dependencies. Extensive experiments on both synthetic and real-world datasets demonstrate that the proposed GiFlow outperforms the state-of-the-art approaches in spatiotemporal imputation. The code is available at https://github.com/zepengzhang/GiFlow.

Problem

• spatiotemporal imputation
• missing data
• graph-informed prior
• flow matching
• error propagation

Innovation

• Graph-Informed Prior
• Flow Matching
• Spatiotemporal Imputation
• Hybrid Vector Field
• Attention Mechanism