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 - 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

All Science Journal Classification (ASJC) codes

  • Computer Vision and Pattern Recognition

Fingerprint Dive into the research topics of 'Learning non-linear dynamical systems by alignment of local linear models'. Together they form a unique fingerprint.

Cite this