An Efficient and Interpretable Autoregressive Model for High-Dimensional Tensor-Valued Time Series

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Cai, Yuxi, Li, Lan, Wang, Yize, Li, Guodong
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866918042918715392
author Cai, Yuxi
Li, Lan
Wang, Yize
Li, Guodong
author_facet Cai, Yuxi
Li, Lan
Wang, Yize
Li, Guodong
contents In autoregressive modeling for tensor-valued time series, Tucker decomposition, when applied to the coefficient tensor, provides a clear interpretation of supervised factor modeling but loses its efficiency rapidly with increasing tensor order. Conversely, canonical polyadic (CP) decomposition maintains efficiency but lacks a precise statistical interpretation. To attain both interpretability and powerful dimension reduction, this paper proposes a novel approach under the supervised factor modeling paradigm, which first uses CP decomposition to extract response and covariate features separately and then regresses response features on covariate ones. This leads to a new CP-based low-rank structure for the coefficient tensor. Furthermore, to address heterogeneous signals or potential model misspecifications arising from stringent low-rank assumptions, a low-rank plus sparse model is introduced by incorporating an additional sparse coefficient tensor. Nonasymptotic properties are established for the ordinary least squares estimators, and an alternating least squares algorithm is introduced for optimization. Theoretical properties of the proposed methodology are validated by simulation studies, and its enhanced prediction performance and interpretability are demonstrated by the El Ni$\tilde{\text{n}}$o-Southern Oscillation example.
format Preprint
id arxiv_https___arxiv_org_abs_2506_01658
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle An Efficient and Interpretable Autoregressive Model for High-Dimensional Tensor-Valued Time Series
Cai, Yuxi
Li, Lan
Wang, Yize
Li, Guodong
Methodology
In autoregressive modeling for tensor-valued time series, Tucker decomposition, when applied to the coefficient tensor, provides a clear interpretation of supervised factor modeling but loses its efficiency rapidly with increasing tensor order. Conversely, canonical polyadic (CP) decomposition maintains efficiency but lacks a precise statistical interpretation. To attain both interpretability and powerful dimension reduction, this paper proposes a novel approach under the supervised factor modeling paradigm, which first uses CP decomposition to extract response and covariate features separately and then regresses response features on covariate ones. This leads to a new CP-based low-rank structure for the coefficient tensor. Furthermore, to address heterogeneous signals or potential model misspecifications arising from stringent low-rank assumptions, a low-rank plus sparse model is introduced by incorporating an additional sparse coefficient tensor. Nonasymptotic properties are established for the ordinary least squares estimators, and an alternating least squares algorithm is introduced for optimization. Theoretical properties of the proposed methodology are validated by simulation studies, and its enhanced prediction performance and interpretability are demonstrated by the El Ni$\tilde{\text{n}}$o-Southern Oscillation example.
title An Efficient and Interpretable Autoregressive Model for High-Dimensional Tensor-Valued Time Series
topic Methodology
url https://arxiv.org/abs/2506.01658