A new FNRS’s Research Fellow for Gérôme Andry in order to overcoming model misspecification with deep learning
Academic Journey
My academic journey began in 2018, initially driven by a general interest in mathematics and the sciences. Quickly, a genuine passion formed around computer science applied to the sciences. Throughout my education, I had multiple opportunities to take on the role of a student instructor, allowing me to share my knowledge and further fuel my enthusiasm for the academic world. In 2023, I earned my Master's degree in Electrical Engineering with a specialization in "Signal Processing and Intelligent Robotics," graduating Magna Cum Laude. My thesis focused on artificial intelligence, particularly the use of deep learning to address data assimilation problems. This work was conducted under the guidance of Professor Gilles Louppe, who is now my thesis supervisor.
Research Project
In the scientific domain, it is common to assume that models perfectly represent reality. However, in practice, these models may diverge from reality or only represent it in highly specific contexts. In some scientific fields, there may not even be appropriate models for certain phenomena. To address this issue, my research project is divided into three distinct parts :
First Part : We will explore physic-informed deep learning techniques which aims to incorporate physical knowledge into the model estimation process to guide inference. In addition, we will dive into the fast moving field of Large Language Models. The purpose being to use them in the context of symbolic regression, with the goal of automating the discovery of laws governing scientific phenomena. Our principal goal will be introducing an uncertainty component into the model inference process.
Second Part : My objective is to make the inference robust against model misspecification. I will study the potential impact of poorly specified models on the inference process and identify avenues for improvement.
Third Part : To apply the developed methods, I plan to undertake an eight-month research internship at Namid Stillman's laboratory at University College of London (UCL). My case study will focus on the collective migration of cells observed in vitro. This field currently lacks robust models to explain cellular dynamics, and existing models are known to be poorly specified.
My research work aims to push the boundaries of model discovery by harnessing deep learning to address model specification issues. These advancements could have diverse applications, ranging from meteorology to robotics, in the realms of science and engineering.
