Exceptional seminar – Cristobal Guzman, 21 April 2022

Hi! PARIS is pleased to propose an exceptional seminar of Cristobal Guzman.

Thanks to Aymeric Dieuleveut, Hi! PARIS Fellowship holder 2021 and Assistant Professor in Statistics at École polytechnique, we have the pleasure to welcome Cristobal Guzman for an exceptional seminar, entitled “Non-Euclidean Differentially Private Stochastic Convex Optimization”.

Thursday 21 April 2022, 11.00am – 1.00pm
Telecom Paris, Room 0A128

On site + Zoom (registration link)

Non-Euclidean Differentially Private Stochastic Convex Optimization

Ensuring privacy of users’ data in machine learning models has become a crucial requirement in multiple domains. In this respect, differential privacy (DP) is the gold standard, due to its general and rigorous privacy guarantees, as well as its high composability. For the particular case of stochastic convex optimization (SCO), recent efforts have established optimal rates for the excess risk under differential privacy in Euclidean setups. These bounds suffer a polynomial degradation of accuracy with respect to the dimension, which limits their applicability in high-dimensional settings. In this talk, I will present nearly-dimension independent rates on the excess risk for DP-SCO in the $\ell_1$ setup, as well as the investigation of more general $\ell_p$ setups, where $1\leq p\leq \infty$.

Based on joint work with Raef Bassily, Michael Menart and Anupama Nandi.


  • Watch the replay (to come)
Cristobal Guzman 

Cristobal Guzman is an Assistant Professor of the Statistics Group in the Department of Applied Mathematics at the University of Twente (on leave from the Catholic University of Chile).  He obtained a Ph.D. in Algorithms, Combinatorics and Optimization at Georgia Tech, under the supervision of Arkadi Nemirovski and Sebastian Pokutta. After graduating, he was a postdoc in the Networks & Optimization Group at CWI-Amsterdam.

He is an expert on optimization, especially stochastic algorithms and iterative methods,  algorithmic stability and differential privacy.

Personal webpage

Google scholar profile