Theory of deep learning
Talks
- M. Wyart, Learning hierarchical representations with deep architectures, Physics of Learning Collaboration, September 2024.
- M. Wyart, On the learnability of hierarchical data, Rome Centre on Mathematics for Modelling and Data Sciences, September 2023.
- F. Cagnetta, Learning hierarchical compositionality with deep convolutional networks: insights from a Random Hierarchy Model, Youth in High-Dimensions, ICTP, May 2023.
- M. Wyart, Loss landscape, over-parametrisation, and curse of dimensionality in deep learning, Ecole de Physique des Houches, July 2022.
- F. Cagnetta, Structure beyond symmetry: locality and compositionality in deep CNNs, Swiss Equivariant Learning Workshop, July 2022.
- M. Wyart, Landscape and learning regimes in deep learning, Glassy Systems and Inter-Disciplinary Applications, Cargese, June 2021.
- M. Wyart, A phase diagram for deep learning unifying jamming, feature learning and lazy training, Workshop on the Theory of Overparameterized Machine Learning (TOPML), Rice University, April 2021.
- M. Geiger, Euclidean equivariant neural networks, Ecole de Physique des Houches, August 2020.
- L. Petrini, Compressing invariant manifolds in neural networks, Ecole de Physique des Houches, August 2020.
- M. Wyart, Loss landscape and performance in deep learning (part 1) and (part 2), Ecole de Physique des Houches, August 2020.
- S. Spigler, Loss landscape and performance in deep learning, Statistical physics of machine learning, ICTS Bangalore, January 2020.
- M. Geiger, Feature and lazy training, Statistical physics of machine learning, ICTS Bangalore, January 2020.
- M. Wyart, Neural Tangent Kernel and over-parametrisation in deep learning, The rough high-dimensional landscape problem, KITP, January 2019.
Lightning Talks
- U. Tomasini, Failure and success of the spectral bias prediction for Laplace Kernel Ridge Regression: the case of low-dimensional data, ICML, July 2022.
- A. Favero, Locality defeats the curse of dimensionality in convolutional teacher-student scenarios, NeurIPS, December 2021.
- L. Petrini, Relative stability toward diffeomorphisms indicates performance in deep nets, NeurIPS, December 2021.
Physics of disordered and glassy systems
- M. Wyart, On the nature of the glass transition, Département de physique de l’ENS, November 2022.
-
E. Agoritsas, Mean-field dynamics of many-body systems: global shear vs random local forcing, Glassy Systems and Inter-Disciplinary Applications, Cargese, June 2021.
-
T. de Geus, Does inertia induce stick-slip friction?, Ecole de Physique des Houches, February 2019.
-
M. Wyart, Theory for swap acceleration near the glass and jamming transition, Disordered serendipity: a glassy path to discovery, Sapienza Università di Roma, September 2018.
-
M. Wyart, Elementary Excitations in Amorphous Materials, Simons Collaboration on Cracking the Glass Problem, February 2017.