Identifying the ultrametricity of time series

Research output: Contribution to journalArticle

26 Citations (Scopus)

Abstract

High dimensional, sparsely populated data spaces have been characterized in terms of ultrametric topology. There are natural, not necessarily unique, tree or hierarchy structures defined by the ultrametric topology. Once such a structure is known, and can be defined, there are various implications including the feasibility of improved computational complexity for operations such as nearest neighbor searching. In this work, we consider the case where the data under investigation is temporal data, in the form of a time series. We develop an approach to characterizing how well time series data can be embedded in an ultrametric topology. Possible applications of this work include: (i) unique fingerprinting of a time series; (ii) discriminating between time series from various domains; and (iii) if data are inherently hierarchical, then using such hierarchies to model and predict.

LanguageEnglish
Pages573-579
Number of pages7
JournalEuropean Physical Journal B
Volume43
Issue number4
DOIs
Publication statusPublished - 1 Feb 2005
Externally publishedYes

Fingerprint

Time series
topology
Topology
hierarchies
Trees (mathematics)
Computational complexity

Cite this

@article{7e78e75975b4443f97b5f82891113222,
title = "Identifying the ultrametricity of time series",
abstract = "High dimensional, sparsely populated data spaces have been characterized in terms of ultrametric topology. There are natural, not necessarily unique, tree or hierarchy structures defined by the ultrametric topology. Once such a structure is known, and can be defined, there are various implications including the feasibility of improved computational complexity for operations such as nearest neighbor searching. In this work, we consider the case where the data under investigation is temporal data, in the form of a time series. We develop an approach to characterizing how well time series data can be embedded in an ultrametric topology. Possible applications of this work include: (i) unique fingerprinting of a time series; (ii) discriminating between time series from various domains; and (iii) if data are inherently hierarchical, then using such hierarchies to model and predict.",
keywords = "Spectroscopy, Neural Network, time series, state physics, Complex systems",
author = "F. Murtagh",
year = "2005",
month = "2",
day = "1",
doi = "10.1140/epjb/e2005-00092-8",
language = "English",
volume = "43",
pages = "573--579",
journal = "European Physical Journal B",
issn = "1434-6028",
publisher = "Springer New York",
number = "4",

}

Identifying the ultrametricity of time series. / Murtagh, F.

In: European Physical Journal B, Vol. 43, No. 4, 01.02.2005, p. 573-579.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Identifying the ultrametricity of time series

AU - Murtagh, F.

PY - 2005/2/1

Y1 - 2005/2/1

N2 - High dimensional, sparsely populated data spaces have been characterized in terms of ultrametric topology. There are natural, not necessarily unique, tree or hierarchy structures defined by the ultrametric topology. Once such a structure is known, and can be defined, there are various implications including the feasibility of improved computational complexity for operations such as nearest neighbor searching. In this work, we consider the case where the data under investigation is temporal data, in the form of a time series. We develop an approach to characterizing how well time series data can be embedded in an ultrametric topology. Possible applications of this work include: (i) unique fingerprinting of a time series; (ii) discriminating between time series from various domains; and (iii) if data are inherently hierarchical, then using such hierarchies to model and predict.

AB - High dimensional, sparsely populated data spaces have been characterized in terms of ultrametric topology. There are natural, not necessarily unique, tree or hierarchy structures defined by the ultrametric topology. Once such a structure is known, and can be defined, there are various implications including the feasibility of improved computational complexity for operations such as nearest neighbor searching. In this work, we consider the case where the data under investigation is temporal data, in the form of a time series. We develop an approach to characterizing how well time series data can be embedded in an ultrametric topology. Possible applications of this work include: (i) unique fingerprinting of a time series; (ii) discriminating between time series from various domains; and (iii) if data are inherently hierarchical, then using such hierarchies to model and predict.

KW - Spectroscopy

KW - Neural Network

KW - time series

KW - state physics

KW - Complex systems

UR - http://www.scopus.com/inward/record.url?scp=17544369448&partnerID=8YFLogxK

U2 - 10.1140/epjb/e2005-00092-8

DO - 10.1140/epjb/e2005-00092-8

M3 - Article

VL - 43

SP - 573

EP - 579

JO - European Physical Journal B

T2 - European Physical Journal B

JF - European Physical Journal B

SN - 1434-6028

IS - 4

ER -