EN FR
EN FR
MODAL - 2012


Section: New Results

Mixture of Gaussians with Missing Data

Participants : Christophe Biernacki, Vincent Vandewalle.

The generative models allow to handle with missing data. This can be easily performed by using the EM algorithm, which has a closed form M-step in the Gaussian setting. This can for instance be useful for distance estimation with missing data. It has been proposed in [18] to improve the distance estimation by fitting a mixture of Gaussian distribution instead of a considering only one Gaussian component. An extension of the previous work including the high setting has been submitted in Neurocomputing journal. This is a joined work with Emil Eirola and Amaury Lendrasse .

A parallel work is in progress on the mixture degeneracy when considering mixture of Gaussians with missing data. It have been experimentally noticed that the degeneracy in this case is particularly slow. This behaviour is different from the usual setting of degeneracy with mixture of Gaussians which is usually rather fast. We are working on the theoretical characterization of this behaviour around a degenerated solution.