I am currently Data Researcher at Ericsson R&D Center France. I am also a reviewer at the following machine learning international conferences: NeurIPS, ICML and ICLR.

Before that, I was a Senior Data Scientist and Researcher for >2 years at MyDataModels, a startup that proposes a WebApp of Data Analytics and Machine Learning, and an R&D engineer at Engie Ineo for 4 years. I also did a postdoc in Statistics and Image Processing sponsored by CNES and l’INRIA with Josiane Zerubia, as well as a 6-month postdoc in Statistics at the University of Tokyo (Japan) with Yuzo Maruyama. I obtained my PhD at Université de Rouen (France), under the supervision of Stéphane Canu and Dominique Fourdrinier.

My topics of interests are the following:

  • sparse optimization methods, both convex (Lasso-type) and nonconvex,
  • feature construction with symbolic regression.
  • model selection – Cp, AIC, BIC, loss estimators, etc.
  • automated machine learning.

I applied those methods to the following areas:

  • Environment – contamination of the atmosphere.
  • Biomedical – interactions the human brain’s regions.
  • Earth observation – object detection on satellite images.
  • Defence – Recognition of radar and telecom signals.

Course at Statlearn 2022

I’ll be giving a course on the L1-penalty at StatLearn’s spring school, which will be held from 4 to 8 April in Cargèse, France.

Semi-supervised work presented at LION

Mikhail Kamalov presented “Graph diffusion & PCA framework for semi-supervised learning” at the Learning and Intelligent OptimizatioN (LION) conference last week.

Paper accepted to EUSIPCO

Our paper “PaZoe: classifying time series with few labels” just got accepted to EUSIPCO!

POT JMLR

Our paper “POT: Python Optimal Transport” was accepted to the Journal of Machine Learning Research (JMLR).

Paper accepted to GECCO

Our paper “Zoetrope Genetic programming” just got accepted to the Genetic and Evolutionary Computation Conference (GECCO).