LUCA SAGLIETTI
Courses a.y. 2023/2024
30537 FOUNDATIONS OF PHYSICS I
30586 MACHINE LEARNING AND ARTIFICIAL INTELLIGENCE LAB
Courses previous a.y.
I am course director and lecturer in the first physics course encountered in the BAI bachelor program, Foundations of Physics I, which offers a broad introduction to Mechanics, Waves and Fluids, and even Special Relativity and Thermodynamics. While being exposed to the beauty of nature is certainly a plus in this course, the main goal is instead to help the students develop strong problem-solving skills.
In the BEMACS bachelor program, instead, I will teach the Computer programming course. This course takes the students a step further into the realm of coding, focusing on more advanced patterns, on computational efficiency and on the relationship between theory and algorithms.
Biographical note
I am an Assistant Professor in the new Computing Sciences Departments of Bocconi University, Milan. I obtained my PhD in 2018 from the Polytechnic of Turin, and then spent few years abroad, at Microsoft Research in Cambridge (MA, USA), at ENS in Paris and at EPFL in Lausanne.
My work lies at the interface between Machine Learning and Statistical Physics, and aims to develop simple theoretical models that allow better understanding of the complex behavior of learning algorithms.
Research interests
I’m currently interested in a range of supervised/semi-supervised learning paradigms that are gaining growing interest in the ML world. For example ML problems that involve multiple learning stages (transfer learning & continual learning, self-supervised learning, distillation, stochastic optimization), multiple learners (adversarial learning, multi-headed networks) and different data subsets (curriculum learning, ML fairness).
Selected Publications
Large deviations in the perceptron model and consequences for active learning
MACHINE LEARNING: SCIENCE AND TECHNOLOGY., 2021
Probing transfer learning with a model of synthetic correlated datasets
MACHINE LEARNING: SCIENCE AND TECHNOLOGY., 2022
Large deviations of semisupervised learning in the stochastic block model
PHYSICAL REVIEW. E, 2022
An Analytical Theory of Curriculum Learning in Teacher-Student Networks
2021
Large deviations for the perceptron model and consequences for active learning
Proceedings of Machine Learning Research. Vol. 107: Mathematical and Scientific Machine Learning, 20-24 July 2020, Princeton University, Princeton, NJ, USA, 2020
Solvable model for inheriting the regularization through knowledge distillation
Proceedings of Machine Learning Research. Vol 145: 2nd Annual Conference on Mathematical and Scientific Machine Learning, 2021