publications

publications and pre-prints, also see my google scholar profile

2023

  1. PhDThesis
    Deep networks training and generalization: insights from linearization
    Thomas George
    Jun 2023
    PhD thesis
    Website

2022

  1. LazyHasty
    Lazy vs hasty: linearization in deep networks impacts learning schedule based on example difficulty
    Thomas George, Guillaume Lajoie, and Aristide Baratin
    Transactions on Machine Learning Research, Jun 2022
    Code Website

2021

  1. The Dynamics of Functional Diversity throughout Neural Network Training
    Lee Zamparo, Marc-Etienne Brunet, Thomas George, and 2 more authors
    NeurIPS (Workshop), Jun 2021
    PDF
  2. NTKAlign
    Implicit regularization via neural feature alignment
    Aristide Baratin, Thomas George, César Laurent, and 4 more authors
    In International Conference on Artificial Intelligence and Statistics, Jun 2021
    arXiv Website
  3. Continual Learning in Deep Networks: an Analysis of the Last Layer
    Timothée Lesort, Thomas George, and Irina Rish
    ICML (Workshop), Jun 2021
    arXiv Poster

2020

  1. Revisiting loss modelling for unstructured pruning
    César Laurent, Camille Ballas, Thomas George, and 2 more authors
    arXiv preprint arXiv:2006.12279, Jun 2020
    arXiv

2018

  1. An Evaluation of Fisher Approximations Beyond Kronecker Factorization
    César Laurent, Thomas George, Xavier Bouthillier, and 2 more authors
    ICLR (Workshop), Jun 2018
  2. Factorized second order methods in neural networks
    Thomas George
    Jun 2018
    Website
  3. EKFAC
    Fast approximate natural gradient descent in a kronecker factored eigenbasis
    Thomas George, César Laurent, Xavier Bouthillier, and 2 more authors
    Advances in Neural Information Processing Systems, Jun 2018
    arXiv Poster