Usage
EM(data, ncomp, model=c("general", "diagonal", "spherical"), class=FALSE,
thres = 0.1, maxit = NULL, rbic=FALSE, debug=FALSE)Arguments
data
matrix of row-elements.
ncomp
maximal number of components in the GMM. In case of degeneracies, the final model size may be less than ncomp.
model
Hypothesis on the modfel to estimate: "general", "diagonal" or "spherical" covariance matrices.
class
If TRUE, hard allocate elements in the E step (see CEM variant in Biernacki et al.). If FALSE, compute soft responsibilities as in usual EM algorithm.
thres
threshold for lower bound variations between 2 iterations. Convergence is decided if this variation is below thres.
maxit
if NULL, the stopping criterion is related to thres. If not NULL, maxit iterations are performed.
rbic
if FALSE, output BIC criterion associated to the obtained GMM. If TRUE, use a variant that accounts for the dimensionality of the model.
debug
if TRUE, display debug markers.