A reproducing kernel Hilbert space framework for pairwise time series distances

Zhengdong Lu, Todd K. Leen, Yonghong Huang, Deniz Erdogmus

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

12 Citations (Scopus)

Abstract

A good distance measure for time series needs to properly incorporate the temporal structure, and should be applicable to sequences with unequal lengths. In this paper, we propose a distance measure as a principled solution to the two requirements. Unlike the conventional feature vector representation, our approach represents each time series with a summarizing smooth curve in a reproducing kernel Hilbert space (RKHS), and therefore translate the distance between time series into distances between curves. Moreover we propose to learn the kernel of this RKHS from a population of time series with discrete observations using Gaussian process-based non-parametric mixed-effect models. Experiments on two vastly different real-world problems show that the proposed distance measure leads to improved classification accuracy over the conventional distance measures.

Original languageEnglish (US)
Title of host publicationProceedings of the 25th International Conference on Machine Learning
Pages624-631
Number of pages8
StatePublished - 2008
Event25th International Conference on Machine Learning - Helsinki, Finland
Duration: Jul 5 2008Jul 9 2008

Other

Other25th International Conference on Machine Learning
CountryFinland
CityHelsinki
Period7/5/087/9/08

Fingerprint

Hilbert spaces
Time series
Experiments

ASJC Scopus subject areas

  • Artificial Intelligence
  • Human-Computer Interaction
  • Software

Cite this

Lu, Z., Leen, T. K., Huang, Y., & Erdogmus, D. (2008). A reproducing kernel Hilbert space framework for pairwise time series distances. In Proceedings of the 25th International Conference on Machine Learning (pp. 624-631)

A reproducing kernel Hilbert space framework for pairwise time series distances. / Lu, Zhengdong; Leen, Todd K.; Huang, Yonghong; Erdogmus, Deniz.

Proceedings of the 25th International Conference on Machine Learning. 2008. p. 624-631.

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

Lu, Z, Leen, TK, Huang, Y & Erdogmus, D 2008, A reproducing kernel Hilbert space framework for pairwise time series distances. in Proceedings of the 25th International Conference on Machine Learning. pp. 624-631, 25th International Conference on Machine Learning, Helsinki, Finland, 7/5/08.
Lu Z, Leen TK, Huang Y, Erdogmus D. A reproducing kernel Hilbert space framework for pairwise time series distances. In Proceedings of the 25th International Conference on Machine Learning. 2008. p. 624-631
Lu, Zhengdong ; Leen, Todd K. ; Huang, Yonghong ; Erdogmus, Deniz. / A reproducing kernel Hilbert space framework for pairwise time series distances. Proceedings of the 25th International Conference on Machine Learning. 2008. pp. 624-631
@inproceedings{b2256fac71c6471fbc7b883d316dc34c,
title = "A reproducing kernel Hilbert space framework for pairwise time series distances",
abstract = "A good distance measure for time series needs to properly incorporate the temporal structure, and should be applicable to sequences with unequal lengths. In this paper, we propose a distance measure as a principled solution to the two requirements. Unlike the conventional feature vector representation, our approach represents each time series with a summarizing smooth curve in a reproducing kernel Hilbert space (RKHS), and therefore translate the distance between time series into distances between curves. Moreover we propose to learn the kernel of this RKHS from a population of time series with discrete observations using Gaussian process-based non-parametric mixed-effect models. Experiments on two vastly different real-world problems show that the proposed distance measure leads to improved classification accuracy over the conventional distance measures.",
author = "Zhengdong Lu and Leen, {Todd K.} and Yonghong Huang and Deniz Erdogmus",
year = "2008",
language = "English (US)",
isbn = "9781605582054",
pages = "624--631",
booktitle = "Proceedings of the 25th International Conference on Machine Learning",

}

TY - GEN

T1 - A reproducing kernel Hilbert space framework for pairwise time series distances

AU - Lu, Zhengdong

AU - Leen, Todd K.

AU - Huang, Yonghong

AU - Erdogmus, Deniz

PY - 2008

Y1 - 2008

N2 - A good distance measure for time series needs to properly incorporate the temporal structure, and should be applicable to sequences with unequal lengths. In this paper, we propose a distance measure as a principled solution to the two requirements. Unlike the conventional feature vector representation, our approach represents each time series with a summarizing smooth curve in a reproducing kernel Hilbert space (RKHS), and therefore translate the distance between time series into distances between curves. Moreover we propose to learn the kernel of this RKHS from a population of time series with discrete observations using Gaussian process-based non-parametric mixed-effect models. Experiments on two vastly different real-world problems show that the proposed distance measure leads to improved classification accuracy over the conventional distance measures.

AB - A good distance measure for time series needs to properly incorporate the temporal structure, and should be applicable to sequences with unequal lengths. In this paper, we propose a distance measure as a principled solution to the two requirements. Unlike the conventional feature vector representation, our approach represents each time series with a summarizing smooth curve in a reproducing kernel Hilbert space (RKHS), and therefore translate the distance between time series into distances between curves. Moreover we propose to learn the kernel of this RKHS from a population of time series with discrete observations using Gaussian process-based non-parametric mixed-effect models. Experiments on two vastly different real-world problems show that the proposed distance measure leads to improved classification accuracy over the conventional distance measures.

UR - http://www.scopus.com/inward/record.url?scp=56449091021&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=56449091021&partnerID=8YFLogxK

M3 - Conference contribution

SN - 9781605582054

SP - 624

EP - 631

BT - Proceedings of the 25th International Conference on Machine Learning

ER -