Supervised dynamic mode decomposition via multitask learning

Keisuke Fujii, Yoshinobu Kawahara

Research output: Contribution to journalArticle

3 Citations (Scopus)

Abstract

Understanding dynamical systems by extracting spatiotemporal patterns from data is fundamental in a variety of fields of engineering and science. Dynamic mode decomposition (DMD) has recently attracted attention in these fields as a way of obtaining a global modal description of a nonlinear dynamical system from data, without requiring explicit prior knowledge. However, DMD is in principle an unsupervised dimensionality reduction algorithm; it is not endowed with the mechanism to utilize label information even if a set of data with different labels is given. In this paper, we propose the algorithm that incorporates label information into DMD via multitask learning by solving sparse-group Lasso. To this end, we estimate sparse weights over dynamic modes in a label-wise manner by regarding data with different labels as different tasks. Modal descriptions estimated by this approach share a part of the global modes, resulting in the extraction of label-specific and common (or mixed) dynamical structures, which could be useful in understanding mechanisms in the spatiotemporal behavior behind data. We investigate the empirical performance using synthetic and real-world datasets, and validate that our algorithm can extract and visualize common and label-specific spatiotemporal structures.

Original languageEnglish
Pages (from-to)7-13
Number of pages7
JournalPattern Recognition Letters
Volume122
DOIs
Publication statusPublished - May 1 2019
Externally publishedYes

Fingerprint

Labels
Decomposition
Nonlinear dynamical systems
Dynamical systems

All Science Journal Classification (ASJC) codes

  • Software
  • Signal Processing
  • Computer Vision and Pattern Recognition
  • Artificial Intelligence

Cite this

Supervised dynamic mode decomposition via multitask learning. / Fujii, Keisuke; Kawahara, Yoshinobu.

In: Pattern Recognition Letters, Vol. 122, 01.05.2019, p. 7-13.

Research output: Contribution to journalArticle

@article{c9f224f55dae49a48ce5c60ce2b7ed01,
title = "Supervised dynamic mode decomposition via multitask learning",
abstract = "Understanding dynamical systems by extracting spatiotemporal patterns from data is fundamental in a variety of fields of engineering and science. Dynamic mode decomposition (DMD) has recently attracted attention in these fields as a way of obtaining a global modal description of a nonlinear dynamical system from data, without requiring explicit prior knowledge. However, DMD is in principle an unsupervised dimensionality reduction algorithm; it is not endowed with the mechanism to utilize label information even if a set of data with different labels is given. In this paper, we propose the algorithm that incorporates label information into DMD via multitask learning by solving sparse-group Lasso. To this end, we estimate sparse weights over dynamic modes in a label-wise manner by regarding data with different labels as different tasks. Modal descriptions estimated by this approach share a part of the global modes, resulting in the extraction of label-specific and common (or mixed) dynamical structures, which could be useful in understanding mechanisms in the spatiotemporal behavior behind data. We investigate the empirical performance using synthetic and real-world datasets, and validate that our algorithm can extract and visualize common and label-specific spatiotemporal structures.",
author = "Keisuke Fujii and Yoshinobu Kawahara",
year = "2019",
month = "5",
day = "1",
doi = "10.1016/j.patrec.2019.02.010",
language = "English",
volume = "122",
pages = "7--13",
journal = "Pattern Recognition Letters",
issn = "0167-8655",
publisher = "Elsevier",

}

TY - JOUR

T1 - Supervised dynamic mode decomposition via multitask learning

AU - Fujii, Keisuke

AU - Kawahara, Yoshinobu

PY - 2019/5/1

Y1 - 2019/5/1

N2 - Understanding dynamical systems by extracting spatiotemporal patterns from data is fundamental in a variety of fields of engineering and science. Dynamic mode decomposition (DMD) has recently attracted attention in these fields as a way of obtaining a global modal description of a nonlinear dynamical system from data, without requiring explicit prior knowledge. However, DMD is in principle an unsupervised dimensionality reduction algorithm; it is not endowed with the mechanism to utilize label information even if a set of data with different labels is given. In this paper, we propose the algorithm that incorporates label information into DMD via multitask learning by solving sparse-group Lasso. To this end, we estimate sparse weights over dynamic modes in a label-wise manner by regarding data with different labels as different tasks. Modal descriptions estimated by this approach share a part of the global modes, resulting in the extraction of label-specific and common (or mixed) dynamical structures, which could be useful in understanding mechanisms in the spatiotemporal behavior behind data. We investigate the empirical performance using synthetic and real-world datasets, and validate that our algorithm can extract and visualize common and label-specific spatiotemporal structures.

AB - Understanding dynamical systems by extracting spatiotemporal patterns from data is fundamental in a variety of fields of engineering and science. Dynamic mode decomposition (DMD) has recently attracted attention in these fields as a way of obtaining a global modal description of a nonlinear dynamical system from data, without requiring explicit prior knowledge. However, DMD is in principle an unsupervised dimensionality reduction algorithm; it is not endowed with the mechanism to utilize label information even if a set of data with different labels is given. In this paper, we propose the algorithm that incorporates label information into DMD via multitask learning by solving sparse-group Lasso. To this end, we estimate sparse weights over dynamic modes in a label-wise manner by regarding data with different labels as different tasks. Modal descriptions estimated by this approach share a part of the global modes, resulting in the extraction of label-specific and common (or mixed) dynamical structures, which could be useful in understanding mechanisms in the spatiotemporal behavior behind data. We investigate the empirical performance using synthetic and real-world datasets, and validate that our algorithm can extract and visualize common and label-specific spatiotemporal structures.

UR - http://www.scopus.com/inward/record.url?scp=85061388340&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85061388340&partnerID=8YFLogxK

U2 - 10.1016/j.patrec.2019.02.010

DO - 10.1016/j.patrec.2019.02.010

M3 - Article

AN - SCOPUS:85061388340

VL - 122

SP - 7

EP - 13

JO - Pattern Recognition Letters

JF - Pattern Recognition Letters

SN - 0167-8655

ER -