Multiclass Gaussianization
Abstract
The Gaussian assumption refers to the use of a Gaussian model to describe a random variable. This assumption is often made in practice, being justified by the central limit theorem. The Gaussian distribution has a set of appealing characteristics related to its mathematical tractability. Therefore, the Gaussian assumption is made actually more often than needed, and a lot of effort has been put into developing methods optimally suited to such environments. However, the Gaussian assumption does not always hold.
Project Goal
We concentrate on Gaussianization for pattern recognition applications. We propose means of adapting classification problems to the Gaussian setup. Therefore, we develop methods to transform the original feature space such that the class-conditional densities are Gaussian.
Project Members
Software
Download: Our elastic-transform based multiclass Gaussianization software (MATLAB) can be downloaded here.
