Learning non-linear dynamical systems by alignment of local linear models

Masao Joko, Yoshinobu Kawahara, Takehisa Yairi

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Learning dynamical systems is one of the important problems in many fields. In this paper, we present an algorithm for learning non-linear dynamical systems which works by aligning local linear models, based on a probabilistic formulation of subspace identification. Because the procedure for constructing a state sequence in subspace identification can be interpreted as the CCA between past and future observation sequences, we can derive a latent variable representation for this problem. Therefore, as in a similar manner to the recent works on learning a mixture of probabilistic models, we obtain a framework for constructing a state space by aligning local linear coordinates. This leads to a prominent algorithm for learning non-linear dynamical systems. Finally, we apply our method to motion capture data and show how our algorithm works well.

Original languageEnglish
Title of host publicationProceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010
Pages1084-1087
Number of pages4
DOIs
Publication statusPublished - Nov 18 2010
Externally publishedYes
Event2010 20th International Conference on Pattern Recognition, ICPR 2010 - Istanbul, Turkey
Duration: Aug 23 2010Aug 26 2010

Publication series

NameProceedings - International Conference on Pattern Recognition
ISSN (Print)1051-4651

Other

Other2010 20th International Conference on Pattern Recognition, ICPR 2010
CountryTurkey
CityIstanbul
Period8/23/108/26/10

Fingerprint

Nonlinear dynamical systems
Data acquisition
Identification (control systems)
Dynamical systems

All Science Journal Classification (ASJC) codes

  • Computer Vision and Pattern Recognition

Cite this

Joko, M., Kawahara, Y., & Yairi, T. (2010). Learning non-linear dynamical systems by alignment of local linear models. In Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010 (pp. 1084-1087). [5595865] (Proceedings - International Conference on Pattern Recognition). https://doi.org/10.1109/ICPR.2010.271

Learning non-linear dynamical systems by alignment of local linear models. / Joko, Masao; Kawahara, Yoshinobu; Yairi, Takehisa.

Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010. 2010. p. 1084-1087 5595865 (Proceedings - International Conference on Pattern Recognition).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Joko, M, Kawahara, Y & Yairi, T 2010, Learning non-linear dynamical systems by alignment of local linear models. in Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010., 5595865, Proceedings - International Conference on Pattern Recognition, pp. 1084-1087, 2010 20th International Conference on Pattern Recognition, ICPR 2010, Istanbul, Turkey, 8/23/10. https://doi.org/10.1109/ICPR.2010.271
Joko M, Kawahara Y, Yairi T. Learning non-linear dynamical systems by alignment of local linear models. In Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010. 2010. p. 1084-1087. 5595865. (Proceedings - International Conference on Pattern Recognition). https://doi.org/10.1109/ICPR.2010.271
Joko, Masao ; Kawahara, Yoshinobu ; Yairi, Takehisa. / Learning non-linear dynamical systems by alignment of local linear models. Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010. 2010. pp. 1084-1087 (Proceedings - International Conference on Pattern Recognition).
@inproceedings{fecdad220d05423980dd1088730c1674,
title = "Learning non-linear dynamical systems by alignment of local linear models",
abstract = "Learning dynamical systems is one of the important problems in many fields. In this paper, we present an algorithm for learning non-linear dynamical systems which works by aligning local linear models, based on a probabilistic formulation of subspace identification. Because the procedure for constructing a state sequence in subspace identification can be interpreted as the CCA between past and future observation sequences, we can derive a latent variable representation for this problem. Therefore, as in a similar manner to the recent works on learning a mixture of probabilistic models, we obtain a framework for constructing a state space by aligning local linear coordinates. This leads to a prominent algorithm for learning non-linear dynamical systems. Finally, we apply our method to motion capture data and show how our algorithm works well.",
author = "Masao Joko and Yoshinobu Kawahara and Takehisa Yairi",
year = "2010",
month = "11",
day = "18",
doi = "10.1109/ICPR.2010.271",
language = "English",
isbn = "9780769541099",
series = "Proceedings - International Conference on Pattern Recognition",
pages = "1084--1087",
booktitle = "Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010",

}

TY - GEN

T1 - Learning non-linear dynamical systems by alignment of local linear models

AU - Joko, Masao

AU - Kawahara, Yoshinobu

AU - Yairi, Takehisa

PY - 2010/11/18

Y1 - 2010/11/18

N2 - Learning dynamical systems is one of the important problems in many fields. In this paper, we present an algorithm for learning non-linear dynamical systems which works by aligning local linear models, based on a probabilistic formulation of subspace identification. Because the procedure for constructing a state sequence in subspace identification can be interpreted as the CCA between past and future observation sequences, we can derive a latent variable representation for this problem. Therefore, as in a similar manner to the recent works on learning a mixture of probabilistic models, we obtain a framework for constructing a state space by aligning local linear coordinates. This leads to a prominent algorithm for learning non-linear dynamical systems. Finally, we apply our method to motion capture data and show how our algorithm works well.

AB - Learning dynamical systems is one of the important problems in many fields. In this paper, we present an algorithm for learning non-linear dynamical systems which works by aligning local linear models, based on a probabilistic formulation of subspace identification. Because the procedure for constructing a state sequence in subspace identification can be interpreted as the CCA between past and future observation sequences, we can derive a latent variable representation for this problem. Therefore, as in a similar manner to the recent works on learning a mixture of probabilistic models, we obtain a framework for constructing a state space by aligning local linear coordinates. This leads to a prominent algorithm for learning non-linear dynamical systems. Finally, we apply our method to motion capture data and show how our algorithm works well.

UR - http://www.scopus.com/inward/record.url?scp=78149472826&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78149472826&partnerID=8YFLogxK

U2 - 10.1109/ICPR.2010.271

DO - 10.1109/ICPR.2010.271

M3 - Conference contribution

AN - SCOPUS:78149472826

SN - 9780769541099

T3 - Proceedings - International Conference on Pattern Recognition

SP - 1084

EP - 1087

BT - Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010

ER -