Accession Number : ADA299178

Title :   Derivation of Joint Representation Mixture Model Equations,

Corporate Author : NAVAL SURFACE WARFARE CENTER DAHLGREN DIV VA

Personal Author(s) : Rogers, George W. ; Lorey, Richard

PDF Url : ADA299178

Report Date : MAY 1995

Pagination or Media Count : 21

Abstract : One of the problems that arises in many large-scale applications of mixture models to density estimation is that, as the size of the data set increases, the class labeled data becomes a (proper) subset of the total data set. That is, while many small data sets may have all the observations labeled as to class membership, large data sets often consist of labeled subsets plus a potentially large unlabeled subset. Thus, it is desirable to have a unified framework for handling this combined supervised (class labeled data)/unsupervised (unlabeled data) problem. This is the motivation behind the following development of joint representation mixture models. The joint representation mixture model is defined, likelihood functions corresponding to different levels of data categorization with respect to class are presented, and the resultant iterative Expectation-Maximization equations are derived. (AN)

Descriptors :   *MATHEMATICAL MODELS, *PROBABILITY DENSITY FUNCTIONS, DATA BASES, ALGORITHMS, PARAMETRIC ANALYSIS, MAXIMUM LIKELIHOOD ESTIMATION, MULTIVARIATE ANALYSIS, STATISTICAL DATA, APPROXIMATION(MATHEMATICS), LOGARITHM FUNCTIONS, NORMAL DISTRIBUTION.

Subject Categories : Statistics and Probability
      Operations Research

Distribution Statement : APPROVED FOR PUBLIC RELEASE