# Reversible Residual Normalization Alleviates Spatio-Temporal Distribution Shift

Source: [Reversible Residual Normalization Alleviates Spatio-Temporal Distribution Shift](<https://www.sota2.com/research/paper/69e5b2df7d0172140ec7de29>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-20T05:00:13.953Z

Authors: Zhaobo Hu, Vincent Gauthier, Mehdi Naima

Published: 2026-04-17T08:40:28.000Z

## Abstract

Distribution shift severely degrades the performance of deep forecasting models. While this issue is well-studied for individual time series, it remains a significant challenge in the spatio-temporal domain. Effective solutions like instance normalization and its variants can mitigate temporal shifts by standardizing statistics. However, distribution shift on a graph is far more complex, involving not only the drift of individual node series but also heterogeneity across the spatial network where different nodes exhibit distinct statistical properties. To tackle this problem, we propose Reversible Residual Normalization (RRN), a novel framework that performs spatially-aware invertible transformations to address distribution shift in both spatial and temporal dimensions. Our approach integrates graph convolutional operations within invertible residual blocks, enabling adaptive normalization that respects the underlying graph structure while maintaining reversibility. By combining Center Normalization with spectral-constrained graph neural networks, our method captures and normalizes complex Spatio-Temporal relationships in a data-driven manner. The bidirectional nature of our framework allows models to learn in a normalized latent space and recover original distributional properties through inverse transformation, offering a robust and model-agnostic solution for forecasting on dynamic spatio-temporal systems.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Spatio-temporal forecasting on SDWPF (20% test)](<https://www.sota2.com/research/sota/spatio-temporal-forecasting-on-sdwpf-20-percent-test.md>): Spatio-temporal forecasting; SDWPF (20% test); MAE: 54.2.

- [Spatio-temporal forecasting on METR-LA (20% test)](<https://www.sota2.com/research/sota/spatio-temporal-forecasting-on-metr-la-20-percent-test.md>): Spatio-temporal forecasting; METR-LA (20% test); MAE: 2.73.

- [Spatio-temporal forecasting on PEMS-BAY (20% test)](<https://www.sota2.com/research/sota/spatio-temporal-forecasting-on-pems-bay-20-percent-test.md>): Spatio-temporal forecasting; PEMS-BAY (20% test); MAE: 1.28.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2604.15838>)

[PDF](<https://arxiv.org/pdf/2604.15838.pdf>)