publications

publications by categories in reversed chronological order. * indicates first-coauthorship.

2022

  1. How deep convolutional neural networks lose spatial information with training
    Umberto M Tomasini *, Leonardo Petrini *, Francesco Cagnetta, and 1 more author
    arXiv preprint arXiv:2210.01506 2022
  2. Learning sparse features can lead to overfitting in neural networks
    Leonardo Petrini *, Francesco Cagnetta *, Eric Vanden-Eijnden, and 1 more author
    Advances in Neural Information Processing Systems 2022

2021

  1. Relative stability toward diffeomorphisms indicates performance in deep nets
    Leonardo PetriniAlessandro FaveroMario Geiger, and 1 more author
    Advances in Neural Information Processing Systems 2021
  2. Landscape and training regimes in deep learning
    Mario GeigerLeonardo Petrini, and Matthieu Wyart
    Physics Reports 2021
  3. Geometric compression of invariant manifolds in neural networks
    Jonas Paccolat, Leonardo PetriniMario Geiger, and 2 more authors
    Journal of Statistical Mechanics: Theory and Experiment 2021