Matteo Vilucchio
PhD Student in Machine Learning and Statistical Physics at EPFL.
About Me
I’m a PhD student at EPFL working on the boundary between High Dimensional Statistics, Computer Science and Statistical Physics. My research focuses on one side to understand the proprites of adversarial/robust estimators in the high dimensional case (where both the number of data used and the dimension of this estimator diverge). On the other side, I am interested in making the non-rigourous but surprisingly effective methods of statistical physics rigorous increasing the toolbox that can be used to study high dimensional problems.
I am lucky to be mentored by Florent Krzakala and to share the lab with all the other IdePHICS memebers.
Outside of my research I spend most of my time climbing either small rocks, medium walls or big mountains. I also enjoy reading and listening to music.
Research Interests
- Mathematical Physics
- Adversarial and Robust Learning
- Statistical Learning Theory
- High Dimensional Statistics
Latest Updates
I will see you at AISTAT2025 in beautiful Phuket, Thailand. I will present our work titled “A High Dimensional Statistical Model for Adversarial Training: Geometry and Trade-offs”.
For the next two weeks I will attend the Winter School Towards a theory for typical-case algorithmic hardness in the beautiful Les Houches, France.
Next week I will be attending the Statistical Physics of Deep Learning school in Como, Italy. I will present our work “A High Dimensional Statistical Model for Adversarial Training: Geometry and Trade-offs” during the poster session. I can’t wait to follow the lectures from Francesca Mastrogiuseppe and Mathieu Wyart!