BDBComp
Parceria:
SBC
Uma Proposta de Melhoria do Algoritmo Guloso de Estimac a o de Mistura de Gaussianas

Andre Paim LemosAntonio Pádua Braga

This work proposes modifications on the stop criterion of the greedy algorithm for Gaussian Mixtures, in order to increase the accuracy in the search for the optimum number of mixture components. In this work, the stop criterion is modified in order to use a sampling multivariate normality test. The algorithm stops when all mixture components pass on the proposed test. The modified algorithm is compared with the original one, that uses parsimony criterion as stop criterion. Numerical simulation results suggest the accuracy improvement when the stop criterion proposed in this work is used.

http://www.lbd.dcc.ufmg.br/colecoes/enia/2011/0014.pdf

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato: bdbcomp@lbd.dcc.ufmg.br
     Mantida por:
LBD