Chasing the Constants and its Implications in Differential Privacy

Explore the deep connections between private continual counting and matrix analysis factorization norms in this Google TechTalk presented by Jalaj Upadhyay. Delve into groundbreaking research that improves upon more than three decades of established results in matrix analysis, with particular focus on establishing upper and lower bounds on factorization norms with an additive gap of $0.14 + o(1)$. Learn how the upper bound is achieved through explicit factorization methods and understand the significant implications these findings have for differential privacy applications. Discover the collaborative research conducted with experts from Google, ISTA, and Fujitsu Research Labs that advances both theoretical understanding and practical applications in privacy-preserving machine learning systems.