Seminario: Differentially-Private Synthetic Data Generation: A Saddle-Point Approach
28 de Agosto, 2025
El Instituto de Ingeniería Matemática y Computacional (IMC) invita al seminario que se dictará el próximo 3 de septiembre.
Cristóbal Guzmán, Instituto de Ingeniería Matemática y Computacional.
Miércoles 3 de septiembre de 2025, 13:40 hrs. (Presencial en auditorio Edificio San Agustín. Link Zoom disponible escribiendo a imc@uc.cl)
ABSTRACT
Generating synthetic data is one of the key problems in private data analysis. In this talk, I will provide a summary of existing work, including well-established privacy attacks from the literature. After this introduction, I will present recent contributions on generating private synthetic data with relative accuracy guarantees (i.e, a mixture of additive and multiplicative error). Our main result provides (counting) error rates can be made poly-logarithmic in the sample size, data universe size, and the number of linear queries; a result which is provably unattainable in the purely-additive counterpart.
BIO
Cristóbal Guzmán is an Associate Professor in Mathematical and Computational Engineering at Pontificia Universidad Católica de Chile. He worked as a Postdoctoral Fellow in the Networks & Optimization group at Centrum Wiskunde & Informatica, Amsterdam. After joining Pontificia Universidad Católica de Chile as Assistant Professor in 2016, he was a Visiting Professor at Centrum Wiskunde & Informatica, and a tenure-track faculty at University of Twente in 2021-22. His research interests include large-scale convex optimization and privacy-preserving machine learning. Cristóbal Guzmán received a Mathematical Engineering degree from Universidad de Chile, Santiago, Chile in 2010, and a Ph.D. in Algorithms, Combinatorics & Optimization from Georgia Institute of Technology, Georgia, USA in 2015. He received a FONDECYT Iniciación grant in 2017, FONDECYT Regular grants in 2021 and 2025, and a Visiting Researcher position at Google in 2023.
