Privacy and trustworthiness of synthetic data: a mathematical framework
by Professor Roman Vershynin, UC Irvine
ABSTRACT: Is there an elegant and general theoretical framework for synthetic data? Can we translate our desiderata - privacy, trustworthiness, fairness, etc. - into rigorous definitions? When can we achieve these desiderata provably? I will discuss two such attempts, one on privacy and the other on trustworthiness of synthetic data. Both attempts lead to concrete open (and fun!) problems in mathematics.
BIO: Roman Vershynin is Professor of Mathematics and Associate Director of the Center for Algorithms, Combinatorics and Optimization at the University of California, Irvine. His research spans high-dimensional probability and mathematical data science, with a particular emphasis on probabilistic structures that appear in random matrix theory, geometric functional analysis, convex and discrete geometry, high-dimensional statistics, information theory, statistical learning, signal processing, numerical analysis, neural networks, and data privacy. He is the author of the book High dimensional probability. An introduction with applications in Data Science, the winner of the 2019 PROSE Award for Mathematics. Dr. Vershynin's honors include an invited talk at the International Congress of Mathematicians in Hyderabad (2010), Bessel Research Award from Humboldt Foundation (2013), Distinguished Mid-Career Faculty Award for Research (2020), and the Medallion Award and Lecture from the Institute of Mathematical Statistics (2022).