Segmental Diffusion: A Distance-Guided Diffusion for Time Series Forecasting
IEEE Signal Processing Letters · 2026 · 33 · 1466-1470
DOI: 10.1109/lsp.2026.3679285 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Signal Processing Letters, 2026. Verfasst von Hao Ren und 2 weiteren Autor:innen. Die Arbeit nennt 25 Literaturangaben. Gefördert von: Science and Technology Planning Project of Liaoning Province.
Abstract (Originalsprache)
Diffusion models have recently shown promise for time series forecasting (TSF), but existing methods rely solely on reverse denoising from Gaussian noise, introducing uncertainty and disrupting temporal structures. We propose Segmental Diffusion (SegDiff), a distance-aware framework for TSF that retrieves semantically similar historical segments to guide each diffusion step via an index-guided conditioning mechanism. A pretrained encoder computes pairwise segment similarities to build the retrieval index. We further introduce a Distance-Aware (DA) Module to estimate transition deltas at each reverse step, explicitly modeling temporal evolution. Trend-aware input augmentation with polynomial regression and moving average smoothing, along with a frequency decomposition module, enhances temporal consistency. Extensive experiments on benchmarks show that SegDiff consistently outperforms existing approaches, validating that segment-level retrieval and distance-aware modeling improve both forecasting accuracy and interpretability.
Förderung
- Science and Technology Planning Project of Liaoning Province
Zitieren
APA
Ren, H., Zhang, L., & Chen, Z. (2026). Segmental Diffusion: A Distance-Guided Diffusion for Time Series Forecasting. IEEE Signal Processing Letters, 33, 1466-1470. https://doi.org/10.1109/lsp.2026.3679285
Vancouver
Ren H, Zhang L, Chen Z. Segmental Diffusion: A Distance-Guided Diffusion for Time Series Forecasting. IEEE Signal Processing Letters. 2026;33:1466-1470. doi:10.1109/lsp.2026.3679285
BibTeX
@article{Ren2026,
author = {Ren, Hao and Zhang, Longxiang and Chen, Zhikui},
title = {Segmental Diffusion: A Distance-Guided Diffusion for Time Series Forecasting},
journal = {IEEE Signal Processing Letters},
year = {2026},
volume = {33},
pages = {1466--1470},
doi = {10.1109/lsp.2026.3679285}
}
EUVIMED