Watch a technical conference talk from FORC 2022 where Mahbod Majid presents groundbreaking research on polynomial-time algorithms for mean estimation with pure differential privacy. Learn about the first efficient method to estimate the mean of a d-variate probability distribution using O(d) independent samples while maintaining privacy guarantees. Explore how the Sum of Squares (SoS) method can be leveraged to design differentially private algorithms, transforming seemingly exponential-time computations into polynomial-time solutions. Discover a novel meta-theorem that demonstrates how low-degree SoS proofs can automatically generate efficient private algorithms, potentially revolutionizing private algorithm design across various applications. The presentation covers introduction to the problem space, detailed algorithm explanations, key challenges, and specific applications in efficient mean estimation.
Efficient Mean Estimation with Pure Differential Privacy via Sum-of-Squares Exponential Mechanism
Harvard CMSA via YouTube
Overview
Syllabus
Introduction
Defining the problem
Algorithms
Problems
Efficient Mean Estimation
Taught by
Harvard CMSA