Francesco Tonin
 |
Francesco Tonin EPFL |
About me
I'm currently a postdoctoral researcher at École Polytechnique Fédérale de Lausanne (EPFL) hosted by Prof. Volkan Cevher.
I obtained a PhD in Machine Learning under the supervision of Prof. Johan Suykens and Prof. Panos Patrinos.
Before that, I obtained my MSc in Computer Science from KU Leuven in 2019 and my BSc in Computer Engineering from Politecnico di Torino in 2017.
More information about my background can be found on my CV.
Latest News
[Jun 2026] Two review papers are accepted on deep learning architectures (link) and kernel SVD (link).
[May 2026] Recognized as Top Reviewer at ICML 2026.
[Mar 2026] MaD-Mix is accepted as Oral presentaion at DATA-FM @ ICLR 2026.
[Feb 2026] Gave a seminar on data mixing in LLMs at ESAT-STADIUS, KU Leuven.
[Jan 2026] Awarded a 50k GPU-hour (GH200) compute grant from Swiss AI to investigate online data pipelines in LLMs.
[Dec 2025] Recognized as Top Reviewer at NeurIPS 2025.
Research Interests
I'm globally interested in Machine Learning and Deep Learning. My research has mainly focused on developing generative kernel-based models that can learn from limited or no labeled data, at the same time making Machine Learning models safer and more interpretable. Currently I'm also interested in the mathematical foundations of Transformers, especially relating to efficient learning.
Specifically, I have focused my attention on:
Dualization in kernel methods and infinite-dimensional feature maps
Energy-based kernel methods for clustering and out-of-distribution detection
Generative Deep Kernel PCA through manifold learning with orthogonality constraints
My list of publications is available here.