Groupe de Travail Analyse, Algorithmique, Apprentissage
Les exposés ont lieu à la salle de séminaire du LJLL (Barre 15-16, 3ème étage, salle 309) Accès
Les organisateurs : Francis, Raphaël, Gérard, Bruno, and Borjan
November 18, 10:30–12:00
Bruno Loureiro (ENS)Title:Learning features with two-layer neural networks, one step at a timeAbstract:Feature learning – or the capacity of neural networks to adapt to the data during training – is often quoted as one of the fundamental reasons behind their unreasonable effectiveness. Yet, making mathematical sense of this seemingly clear intuition is still a largely open question. In this talk, I will discuss a simple setting where we can precisely characterise how features are learned by a two-layer neural network during the very first few steps of training, and how these features are essential for the network to efficiently generalise under limited availability of data.Based on the following works: https://arxiv.org/abs/2305.18270, https://arxiv.org/abs/2402.04980, https://arxiv.org/abs/2410.18938
in room 15-16-309 at Jussieu (salle de séminaire of the LJLL), as usual.
Looking forward to seeing you all!
Best,
the organizers (Francis, Raphaël, Gérard, Bruno, and Borjan)