I am an Assistant Professor in Computer Science and Software/Computer/Electrical Engineering at Université Laval (Québec City, Canada). I am also affiliated with Mila — Quebec Artificial Intelligence Institute (Montreal, Canada). Prior to that I was a postdoctoral researcher with Prof. Joelle Pineau at McGill University (Montreal, Canada).
I completed a PhD in Electrical Engineering (2011-2018) in the Computer Vision and Systems Laboratory at Université Laval, where I was supervised by Prof. Christian Gagné and co-supervised by Prof. Joelle Pineau (McGill University). I also collaborated with Odalric-Ambrym Maillard (INRIA Lille – Nord Europe). Before that I obtained a MSc in Electrical Engineering (2009-2011) and a BSc in Computer Engineering (2004-2009) from Université Laval.
Announcement
I am currently recruiting students! If you would be interested in working with me, have a look at these resources and ping me for a discussion:
- Bandit Algorithms by Tor Lattimore and Csaba Szepesvari
- Algorithms for Reinforcement Learning by Csaba Szepesvari
Research interests
During my Master’s I got really interested by Reinforcement Learning. During my PhD, I have mostly focused my attention on the exploration-exploitation tradeoff through bandits and their application to real-world problems. More specifically, I have worked on:
- Contextual bandits and their application to personalized treatment policy learning and recommendation systems (with Coveo)
- Structured bandits using kernel regression
- Multi-objective bandits and their application to design of experiments for online optimization of super-resolution imaging parameters (with researcher Flavie Lavoie-Cardinal from CERVO and Prof. Paul De Koninck from Université Laval)
- Combinatorial bandits and their application to online feature selection
I am an engineer at the heart. Applied bandits therefore make a lot of sense for me. I have been collaborating with several clinicians/researchers in healthcare for a long time. I think machine learning can bring a lot to health-related research and, reversely, that many really interesting problems for machine learners raise from applications.
