Currently I am a postdoctoral researcher in Machine Learning (and Statistics) at the University of Grenoble within the AMA team of the LIG lab Research. I work with Eric Gaussier, Marianne Clausel and
Emilie Devijver in a partnership with Total, the financier of my postdoc. The research goal is to learn a model allowing predict the welding quality and including heterogeneous input and output variables with uncertain measures. Preliminary work on a real data base (provided by Total) encourages us to focus our research direction on ensemble methods. Then, we are developing and implementing ensemble methods with quadratic and quantile regression which, during the learning task, takes into account distributions of each quantitative observation from the dataset.
Before that I did a PhD in EM simultaneous estimation of a structural equation modeling and its latent factors at the University of Montpellier. Currently, I work on some perspectives of this work with Vincent Brault and Emilie Devijver. We develop a supplementary step of this approach algorithm. The goal is to catch the different explanatory blocks of variables in an unsupervised way. I am also working on the publication of this EM approach as an R package. Among the several perspective of my Ph.D work, causality can provide a great contribution to the EM estimation approach of structural equation modeling. That’s why I am a member of a Causality reading group at LIG.