Discriminant Analysis by Gaussian Mixtures
- 1 January 1996
- journal article
- research article
- Published by Oxford University Press (OUP) in Journal of the Royal Statistical Society Series B: Statistical Methodology
- Vol. 58 (1) , 155-176
- https://doi.org/10.1111/j.2517-6161.1996.tb02073.x
Abstract
SUMMARY: Fisher-Rao linear discriminant analysis (LDA) is a valuable tool for multigroup classification. LDA is equivalent to maximum likelihood classification assuming Gaussian distributions for each class. In this paper, we fit Gaussian mixtures to each class to facilitate effective classification in non-normal settings, especially when the classes are clustered. Low dimensional views are an important by-product of LDA—our new techniques inherit this feature. We can control the within-class spread of the subclass centres relative to the between-class spread. Our technique for fitting these models permits a natural blend with nonparametric versions of LDA.Keywords
This publication has 11 references indexed in Scilit:
- Penalized Discriminant AnalysisThe Annals of Statistics, 1995
- Flexible Discriminant Analysis by Optimal ScoringJournal of the American Statistical Association, 1994
- Neural Networks: A Review from a Statistical PerspectiveStatistical Science, 1994
- Generalized clustering networks and Kohonen's self-organizing schemeIEEE Transactions on Neural Networks, 1993
- Canonical Correlation Analysis When the Data are CurvesJournal of the Royal Statistical Society Series B: Statistical Methodology, 1993
- Discriminant Analysis and Statistical Pattern RecognitionWiley Series in Probability and Statistics, 1992
- Statistical classification using a linear mixture of two multinormal probability densitiesPattern Recognition Letters, 1991
- Multivariate Adaptive Regression SplinesThe Annals of Statistics, 1991
- Self-Organization and Associative MemoryPublished by Springer Nature ,1989
- CANONICAL VARIATE ANALYSIS—A GENERAL MODEL FORMULATIONAustralian Journal of Statistics, 1984