Gaussian Mixture Models Based on Principal Components and Applications

Table 6

Forensic glass fragments data: parameter estimates and BIC of the bivariate GMM (second technique) by taking cluster means as initial values of model means, and the initial values for PC1 were −0.36697 and 1.45934 and those for PC2 are 0.35596 and −1.41558.

Values	Model parameters
	First component					Second component
	Parameter estimation	0.96102	–0.10113	−0.12063	1.10542	0.72645	0.03898	2.49307	2.97325	6.61789	4.94843
Log-likelihood = –563.0958, BIC = 1174.485; Number of iterations: 28