VBmix (version 0.3.2)

mppca:

Description

estimates the variational posterior distribution of a MPPCA on a data set. A lower bound is calculated and monitored at each iteration. This posterior can be used for various purposes (e.g. MC proposal distribution). It can be transformed using mppcaToGmm and subMppca, outputing a GMM.

Usage

mppca(data, ncomp, thres = 0.1, maxit = NULL, qmax = NULL)

Arguments

data

matrix of row-elements.

ncomp

number of components in the posterior.

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.

qmax

maximal rank of the posterior factor matrices. If NULL, is set to d-1.

Value

estimated posterior MPPCA with ncomp components.

References

Beal, M. J. (2003) _Variational Algorithms for approximate inference_, PhD Thesis, University of London.

See Also

newMppca mppcaToGmm subMppca

Examples

# for packaging needs, a low amount of initial components (ie 10) was used.
# A larger amount may be used for better results.
temp <- mppca(pendat, 10, maxit=20, qmax=8)