ABSTRACT
The first instance of Gaussian mixture models being used for clustering is sometimes attributed to J. H. Wolfe. Wolfe uses the idea of a mixture to define a cluster, or type: a type is a distribution which is one of the components of mixture of distributions. Wolfe computes maximum likelihood estimates for Gaussian model-based clustering. Wolfe draws an analogy between his approach for Gaussian mixtures with common covariance matrices and one of the criteria described by H. P. Friedman and J. Rubin. E. L. Scott and M. J. Symons consider approaches and parameter estimation in a Gaussian model-based classification scenario, and point out the analogy of one of these approaches with the work of A. W. F. Edwards and L. L. Cavalli–Sforza. The Gaussian mixture model can be used to effectively cluster and classify data. Clusters are generally taken as synonymous with mixture components and a similar one-to-one relationship is commonly assumed in model-based classification.
