Section: New Software and Platforms

GLLiM

Gaussian Locally Linear Mapping

Keywords: Regression - Machine learning - Gaussian mixture

Scientific Description: GLLiM is a flexible tool for probabilistic non-linear regression using Gaussian mixtures. Using an inverse regression strategy with a reduced number of parameters, it is particularly suited for high- to low-dimensional regression tasks. It also enables the modeling of additional unobserved non-linear effects on input data. The method was published in [Deleforge et al., IJNS 2015]. The toolbox include an example of application to head pose estimation from synthetic images.

• Participant: Antoine Deleforge

• Contact: Antoine Deleforge

• Publication: hal-00863468, version 3

• URL: https://team.inria.fr/perception/gllim_toolbox/