Clustering via the Distribution Mixtures

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21340%2F10%3A00176166" target="_blank" >RIV/68407700:21340/10:00176166 - isvavai.cz</a>

Výsledek na webu

—

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Clustering via the Distribution Mixtures

Popis výsledku v původním jazyce

The finite distribution mixtures present a wide class of probability distributions. Apart from the obvious applications, the mixtures are successfully applied in the model based clustering. If we constraint the members of the mixture to arise from one specific family or type of parametric distributions, each cluster would refer to one component of the mixture. The membership of the observed sample to a cluster is given simply as the maximum probability on the components of the mixture, i.e. by the Mahalanobis distance, and weighted by the weights of the mixture. This approach is feasible even for overlapping clusters and strongly uneven numbers of the members of the clusters, where standard methods of cluster analysis fall short. We focus on the problem of fitting the mixture to observed sample using the maximum likelihood approach and the EM algorithm, as well as the assessment of the optimal number of components.

Název v anglickém jazyce

Clustering via the Distribution Mixtures

Popis výsledku anglicky

The finite distribution mixtures present a wide class of probability distributions. Apart from the obvious applications, the mixtures are successfully applied in the model based clustering. If we constraint the members of the mixture to arise from one specific family or type of parametric distributions, each cluster would refer to one component of the mixture. The membership of the observed sample to a cluster is given simply as the maximum probability on the components of the mixture, i.e. by the Mahalanobis distance, and weighted by the weights of the mixture. This approach is feasible even for overlapping clusters and strongly uneven numbers of the members of the clusters, where standard methods of cluster analysis fall short. We focus on the problem of fitting the mixture to observed sample using the maximum likelihood approach and the EM algorithm, as well as the assessment of the optimal number of components.

Druh

D - Stať ve sborníku

CEP obor

BA - Obecná matematika

OECD FORD obor

—

Projekt

—

Návaznosti

S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2010

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Doktorandské dny 2010

ISBN

978-80-01-04644-9

ISSN

—

e-ISSN

—

Počet stran výsledku

10

Strana od-do

—

Název nakladatele

Česká technika - nakladatelství ČVUT

Místo vydání

Praha

Místo konání akce

Praha

Datum konání akce

19. 11. 2010

Typ akce podle státní příslušnosti

CST - Celostátní akce

Kód UT WoS článku

—