skip to main content

Title: Use of the AIC with the EM algorithm: A demonstration of a probability model selection technique

The problem of discriminating between two potential probability models, a Gaussian distribution and a mixture of Gaussian distributions, is considered. The focus of our interest is a case where the models are potentially non-nested and the parameters of the mixture model are estimated through the EM algorithm. The AIC, which is frequently used as a criterion for discriminating between non-nested models, is modified to work with the EM algorithm and is shown to provide a model selection tool for this situation. A particular problem involving an infinite mixture distribution known as Middleton`s Class A model is used to demonstrate the effectiveness and limitations of this method.
Authors:
;  [1]
  1. Lawrence Livermore National Lab., CA (United States)
Publication Date:
OSTI Identifier:
145998
Report Number(s):
CONF-9411140--Absts.
ON: DE95017252; TRN: 95:007225-0048
Resource Type:
Conference
Resource Relation:
Conference: Imaging sciences workshop, Livermore, CA (United States), 15-16 Nov 1994; Other Information: PBD: 15 Nov 1994; Related Information: Is Part Of Imaging sciences workshop; Candy, J.V.; PB: 101 p.
Research Org:
Lawrence Livermore National Lab., CA (United States)
Country of Publication:
United States
Language:
English
Subject:
99 MATHEMATICS, COMPUTERS, INFORMATION SCIENCE, MANAGEMENT, LAW, MISCELLANEOUS; GAUSS FUNCTION; COMPARATIVE EVALUATIONS; STATISTICAL MODELS; GAUSSIAN PROCESSES; MIXTURES