Simone Bombari

smile_2.jpg

Hi! I am Simone Bombari, and this is my website! :muscle:

I am a PhD student at ISTA, happy and lucky to be advised by Professor Marco Mondelli. Right now, I enjoy thinking about research questions connected with deep learning, high dimensional probability, and trustworthy machine learning. For example, I like to think if over-parameterized models trained on high-dimensional data hide some not-necessarily-trivial behavior, with maybe some implications connected to privacy or robustness.

Before joining ISTA I studied Physics at Scuola Normale Superiore and University of Pisa. During my studies, I ended up focusing on statistical physics and quantum information. Then, I slowly migrated towards statistics and machine learning. :computer:

During my PhD, I had the opportunity to intern as an applied scientist at AWS, and as NLP quant researcher at G Research. Since the fall of 2024, I am supported by a Google PhD fellowship.

I am currently a visiting scholar at the University of Southern California, in the groups of Yizhe Zhu and Mahdi Soltanolkotabi.

Of course, I have many interests beyond my research. I have a deep passion for sports, fitness, games and many other things! :soccer:

selected publications

  1. no_muscle.jpg
    Beyond the Universal Law of Robustness: Sharper Laws for Random Features and Neural Tangent Kernels
    Simone Bombari ,  Shayan Kiyani ,  and  Marco Mondelli
    Proceedings of the 40th International Conference on Machine Learning, 2023
    arXiv
  2. privacy.png
    Privacy for Free in the Overparameterized Regime
    Simone Bombari ,  and  Marco Mondelli
    Proceedings of the National Academy of Sciences, 2025
    arXiv
  3. DP-SGD-ODE_new.png
    Better Rates for Private Linear Regression in the Proportional Regime via Aggressive Clipping
    Simone Bombari ,  Inbar Seroussi ,  and  Marco Mondelli
    arXiv, 2025
    arXiv