Salvato in:
Dettagli Bibliografici
Autori principali: Zhang, Zhiwei, Du, Xinyi, Guo, Xuanchi, Wang, Weihao, Han, Wenjuan
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2511.08396
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911259602976768
author Zhang, Zhiwei
Du, Xinyi
Guo, Xuanchi
Wang, Weihao
Han, Wenjuan
author_facet Zhang, Zhiwei
Du, Xinyi
Guo, Xuanchi
Wang, Weihao
Han, Wenjuan
contents Multivariate time series forecasting is crucial across a wide range of domains. While presenting notable progress for the Transformer architecture, iTransformer still lags behind the latest MLP-based models. We attribute this performance gap to unstable inter-channel relationships. To bridge this gap, we propose EMAformer, a simple yet effective model that enhances the Transformer with an auxiliary embedding suite, akin to armor that reinforces its ability. By introducing three key inductive biases, i.e., \textit{global stability}, \textit{phase sensitivity}, and \textit{cross-axis specificity}, EMAformer unlocks the further potential of the Transformer architecture, achieving state-of-the-art performance on 12 real-world benchmarks and reducing forecasting errors by an average of 2.73\% in MSE and 5.15\% in MAE. This significantly advances the practical applicability of Transformer-based approaches for multivariate time series forecasting. The code is available on https://github.com/PlanckChang/EMAformer.
format Preprint
id arxiv_https___arxiv_org_abs_2511_08396
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle EMAformer: Enhancing Transformer through Embedding Armor for Time Series Forecasting
Zhang, Zhiwei
Du, Xinyi
Guo, Xuanchi
Wang, Weihao
Han, Wenjuan
Machine Learning
Multivariate time series forecasting is crucial across a wide range of domains. While presenting notable progress for the Transformer architecture, iTransformer still lags behind the latest MLP-based models. We attribute this performance gap to unstable inter-channel relationships. To bridge this gap, we propose EMAformer, a simple yet effective model that enhances the Transformer with an auxiliary embedding suite, akin to armor that reinforces its ability. By introducing three key inductive biases, i.e., \textit{global stability}, \textit{phase sensitivity}, and \textit{cross-axis specificity}, EMAformer unlocks the further potential of the Transformer architecture, achieving state-of-the-art performance on 12 real-world benchmarks and reducing forecasting errors by an average of 2.73\% in MSE and 5.15\% in MAE. This significantly advances the practical applicability of Transformer-based approaches for multivariate time series forecasting. The code is available on https://github.com/PlanckChang/EMAformer.
title EMAformer: Enhancing Transformer through Embedding Armor for Time Series Forecasting
topic Machine Learning
url https://arxiv.org/abs/2511.08396