Hi, I’m a French researcher who recently graduated with a PhD in evolutionary machine learning at ISAE-SUPAERO in Toulouse, France. Welcome to my personal web page!

Bio

I received a Masters in Engineering from ISAE-SUPAERO with a specialization in Machine Learning and robotics, and a Masters in Operations Research from the University of Toulouse in 2020.

After a research internship on neuroevolution in the SuReLI group at ISAE-SUPAERO, I continued with a PhD in machine learning in the same lab, focusing on Evolution Strategies to train artificial neural networks for policy search. I graduated on April 22, 2024, with a thesis entitled “Leveraging structures in Evolutionary Neural Policy Search”.

Work overview

The main focus on my PhD research has been evolutionary neural policy search, where evolutionary optimization methods (such as Evolution Strategies) are used to optimize neural networks, that are then used as controllers (policies) in a sequential decision-making problem. This includes tasks such as games, robotics control, or learning to drive a car.

My main research topics have included neural network representations in evolution and their optimization with meta-evolution, resampling methods for EA, or mixing evolution with deep RL.

After visiting Antoine Cully’s lab at Imperial College London in 2023, I studied how policy behavior, used in Quality-Diversity methods, can be leveraged to better search for good policies. We introduced the framework of Quality with Just Enough Diversity (JEDi), which won the Best Paper Award at GECCO 24.

Research interests

While my published research is focused on evolution for neural networks, I had the chance of working alongside researchers focusing on various domains of machine learning. From these interactions, and personal interests, I have also built a wider understanding of policy search through deep RL and Quality-Diversity methods, but also of explainable methods like Genetic Programming or applications like robotics. With an engineering background, I enjoy building tools and have used various parallelization approaches from MPI to Jax, and developped a custom SLURM jobs creation tool to burn CPUs scale research more easily.

I’m always interested in learning new things around ML, so please don’t hesitate to email me to chat!