Barbier group
We are based at ICTP and SISSA in Trieste. Current funding: ERC Starting Grant CHORAL (Computational Hardness Of RepresentAtion Learning).
Principal investigator
-
Jean Barbier
Postdocs
-
Minh-Toan Nguyen
Information theory, random matrix theory and statistical physics for high-dimensional inference.
Background
Engineering degree from École Polytechnique and Master's in "Mathematics of Randomness" at Paris-Saclay (2020). PhD at Université Grenoble Alpes with Romain Couillet, with a thesis on the replica method and asymptotic equivalence.
-
Mauro Pastore
Feature learning in overparametrized networks, its relation to kernel methods, and the role of data structure.
Background
PhD in theoretical physics at the University of Milan with Sergio Caracciolo. Postdocs in Paris at LPTMS with Silvio Franz (Simons collaboration on the glass problem) and at LPENS with Simona Cocco and Rémi Monasson, on inference in biology.
-
Brandon Livio Annesi
Statistical mechanics of learning, random matrix theory and high-dimensional inference.
Background
PhD at Bocconi University (2025) with Carlo Lucibello on statistical physics of non-convex neural network models, then a year at Sapienza University of Rome with Chiara Cammarota on random-matrix approaches to learning.
PhD students
-
Rudy Skerk
Statistical mechanics, information theory and random matrices for optimization, inference and learning.
Background
Bachelor's in physics at the University of Trieste, then the International Master in Physics of Complex Systems (Trieste, Turin, Paris). PhD student in Theoretical and Scientific Data Science at SISSA.
-
Ali Hussaini Umar
Understanding deep learning through its representations and weights.
Background
Bachelor's in mathematics at Yobe State University, Master's at AIMS Ghana and the ICTP diploma in Quantitative Life Sciences. PhD student in Theoretical and Scientific Data Science at SISSA.
-
Cristofer Erazo
Learning dynamics and representations in modern networks, in particular in-context learning in transformers.
Background
Bachelor's in physics at Escuela Politécnica Nacional (Ecuador) and the ICTP diploma in Quantitative Life Sciences, where he received the Qaisar and Monika Shafi Prize (2023–24). PhD student in Theoretical and Scientific Data Science at SISSA.
-
Luca Camagna
Statistical physics and random matrix theory of AI architectures, with a focus on information-theoretic limits.
Background
International Master in Physics of Complex Systems (Trieste, Turin, Paris). PhD student in Theoretical and Scientific Data Science at SISSA.
Former members
| Name | In the group as | Now |
|---|---|---|
| Gibbs Nwemadji | PhD student | Postdoc, Yale University |
| Francesco Camilli | Postdoc | Professor, University of Bologna |
| Daria Tieplova | Postdoc | Postdoc, Aarhus University |
| Anas Rahman | Postdoc | Postdoc, HKUST |
| Manuel Sáenz | Postdoc | Professor, University of San Andrés |
| Eleonora Bergamin | Visiting student | PhD student, SISSA |
| Shu Zhou | Visiting student | Postdoc, SISSA |
| Yizhou Xu | Visiting student | PhD student, EPFL |
| Rodrigo Pérez Ortiz | Visiting student | PhD student, University of Bologna |
| Koki Okajima | Visiting student | Professor, Kyoto University |
| Sébastien Pierre | Visiting student | Student, École Polytechnique |
| Tianqi Hou | Visiting student | Huawei |
Join us
- PhD
- Apply to the PhD program in Theoretical and Scientific Data Science at SISSA, which opens a call every year, and mention the group in your application. Write to me beforehand if you want to discuss projects.
- Master's thesis
- Contact me directly with a CV and a few lines on what you would like to work on.
- Postdoc
- Openings are announced on this site and by ICTP. Spontaneous enquiries are welcome.
- Diploma
- Several group members came through ICTP's one-year pre-PhD Postgraduate Diploma Programme in Quantitative Life Sciences; the diploma in Mathematics is another route into the group. Scholarships are available for students from developing countries.