Is PGMM_Model::loadPGMMfromMATLAB() really efficient?
Previous implementation of this function loaded only one file for the centers of the Gaussians, and another file containing the covariance matrices. The function assumed a specific saving protocol when the files were generated in MATLAB. Now the function loads a different file for every tasl-parameter defined in the problem, which may lead to have too many files. I think the previous implementation was better in this sense. Why was it changed?