Parametric estimation for normal mixtures

Authors:
James C Bezdek;Richard J Hathaway;Vicki J Huggins
Affiliations:
Computer Science Department, University of South Carolina, Columbia, SC 29208, USA;Department of Mathematics and Statistics, University of South Carolina, Columbia, SC 29208, USA;Department of Mathematics and Statistics, University of South Carolina, Columbia, SC 29208, USA
Venue:
Pattern Recognition Letters
Year:
1985

Citing 1
Cited 2

Optimal Fuzzy Partitions: A Heuristic for Estimating the Parameters in a Mixture of Normal Distributions

IEEE Transactions on Computers

Extending fuzzy and probabilistic clustering to very large data sets

Computational Statistics & Data Analysis
A survey of fuzzy clustering

Mathematical and Computer Modelling: An International Journal

Quantified Score

Hi-index	0.10

Visualization

Abstract

Described here are two approaches for estimating the parameters (a-priori probabilities, means, and covariances) of a mixture of normal distributions, given a finite sample X drawn from the mixture. One approach is based on a modification of the EM algorithm for computing maximum-likelihood estimates, while the other makes use of the Fuzzy c-Means algorithms for locating clusters. The reliability, accuracy, and efficiency of these two algorithms are compared using samples drawn from three artificial univariate normal mixtures of two classes.