Joachim Bona-Pellissier

About

I am a postdoctoral researcher at the Machine Learning Genoa Center (MaLGa), where I work under the supervision of Lorenzo Rosasco.

Prior to that, I completed a PhD in applied mathematics in December 2023 in Toulouse, France, under the supervision of Francois Malgouyres and Francois Bachoc.

Research interests:

  • Physics-informed machine learning
  • Deep learning theory
  • Geometry of ReLU networks
  • Kernel methods

Preprints

  1. PIKS: Universal Physics-Informed Kernel Methods

    Joachim Bona-Pellissier, Giacomo Meanti, Matteo Santacesaria, Lorenzo Rosasco. Preprint, 2026.

  1. Geometry-induced Regularization in Deep ReLU Neural Networks

    Joachim Bona-Pellissier, Francois Malgouyres, Francois Bachoc. Preprint, 2024.

Publications

  1. Parameter identifiability of a deep feedforward ReLU neural network

    Joachim Bona-Pellissier, Francois Bachoc, Francois Malgouyres. Machine Learning, 2023.

  2. Local Identifiability of Deep ReLU Neural Networks: the Theory

    Joachim Bona-Pellissier, Francois Malgouyres, Francois Bachoc. NeurIPS 35, 2022.

Talks

  1. 2026
  2. 2026
    InterCoML Seminar, COST Action InterCoML, Online.
  3. 2026
    MATH4AIML - UMI Workshop, Sapienza Universita di Roma, Italy.
  4. 2025
    Representations Neuronales Implicites : des NeRF aux PINN, Institut Henri Poincare, Paris, France.
  5. 2025
    NPTwins, Universita di Messina, Italy.