UniMamba: A Unified Spatial-Temporal Modeling Framework with State-Space and Attention Integration

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Chen, Xingsheng, Mu, Xianpei, Yi, Deyu, Yuan, Yilin, He, Xingwei, Gao, Bo, Zhang, Regina, Lio, Pietro, Yiu, Siu-Ming
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910217962258432
author Chen, Xingsheng
Mu, Xianpei
Yi, Deyu
Yuan, Yilin
He, Xingwei
Gao, Bo
Zhang, Regina
Lio, Pietro
Yiu, Siu-Ming
author_facet Chen, Xingsheng
Mu, Xianpei
Yi, Deyu
Yuan, Yilin
He, Xingwei
Gao, Bo
Zhang, Regina
Lio, Pietro
Yiu, Siu-Ming
contents Multivariate time series forecasting is fundamental to numerous domains such as energy, finance, and environmental monitoring, where complex temporal dependencies and cross-variable interactions pose enduring challenges. Existing Transformer-based methods capture temporal correlations through attention mechanisms but suffer from quadratic computational cost, while state-space models like Mamba achieve efficient long-context modeling yet lack explicit temporal pattern recognition. Therefore we introduce UniMamba, a unified spatial-temporal forecasting framework that integrates efficient state-space dynamics with attention-based dependency learning. UniMamba employs a Mamba Variate-Channel Encoding Layer enhanced with FFT-Laplace Transform and TCN to capture global temporal dependencies, and a Spatial Temporal Attention Layer to jointly model inter-variate correlations and temporal evolution. A Feedforward Temporal Dynamics Layer further fuses continuous and discrete contexts for accurate forecasting. Comprehensive experiments on eight public benchmark datasets demonstrate that UniMamba consistently outperforms state-of-the-art forecasting models in both forecasting accuracy and computational efficiency, establishing a scalable and robust solution for long-sequence multivariate time-series prediction.
format Preprint
id arxiv_https___arxiv_org_abs_2604_16325
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle UniMamba: A Unified Spatial-Temporal Modeling Framework with State-Space and Attention Integration
Chen, Xingsheng
Mu, Xianpei
Yi, Deyu
Yuan, Yilin
He, Xingwei
Gao, Bo
Zhang, Regina
Lio, Pietro
Yiu, Siu-Ming
Machine Learning
Artificial Intelligence
Multivariate time series forecasting is fundamental to numerous domains such as energy, finance, and environmental monitoring, where complex temporal dependencies and cross-variable interactions pose enduring challenges. Existing Transformer-based methods capture temporal correlations through attention mechanisms but suffer from quadratic computational cost, while state-space models like Mamba achieve efficient long-context modeling yet lack explicit temporal pattern recognition. Therefore we introduce UniMamba, a unified spatial-temporal forecasting framework that integrates efficient state-space dynamics with attention-based dependency learning. UniMamba employs a Mamba Variate-Channel Encoding Layer enhanced with FFT-Laplace Transform and TCN to capture global temporal dependencies, and a Spatial Temporal Attention Layer to jointly model inter-variate correlations and temporal evolution. A Feedforward Temporal Dynamics Layer further fuses continuous and discrete contexts for accurate forecasting. Comprehensive experiments on eight public benchmark datasets demonstrate that UniMamba consistently outperforms state-of-the-art forecasting models in both forecasting accuracy and computational efficiency, establishing a scalable and robust solution for long-sequence multivariate time-series prediction.
title UniMamba: A Unified Spatial-Temporal Modeling Framework with State-Space and Attention Integration
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2604.16325