estimates the variational posterior distribution of a MPPCA that aggregates a collection of input MPPCA models. 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. The maximal rank of output factor matrices is determined by the inputs.
Usage
mmppca(mods, ncomp, thres = 0.1, maxit = NULL)
Arguments
mods
input MPPCA that concatenates the set of components to aggregate.
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.
Value
estimated posterior MPPCA with ncomp components.
References
Bruneau, P., Gelgon, M. and Picarougne, F. (2010) _Aggregation of probabilistic PCA mixtures with
a variational-Bayes technique over parameters_, ICPR'10. Bruneau, P., Gelgon, M. and Picarougne, F. (2011) _Component-level aggregation of probabilistic PCA mixtures using variational-Bayes_, Tech Report, http://hal.archives-ouvertes.fr/docs/00/56/72/99/PDF/techrep.pdf.