Explore the cutting-edge realm of private stochastic convex optimization in this 24-minute conference talk presented at the Association for Computing Machinery (ACM). Delve into key concepts such as stochastic convex optimization, private empirical risk minimization, and privacy amplification. Learn about the innovative Snowball SGT algorithm and its applications in reducing sensitivity and iterative localization. Gain insights into strongly convex problems and their implications for privacy-preserving optimization techniques. Discover how these advanced methods can achieve optimal rates in linear time, revolutionizing the field of machine learning and data analysis while maintaining privacy guarantees.
Private Stochastic Convex Optimization: Optimal Rates in Linear Time
Association for Computing Machinery (ACM) via YouTube
Overview
Syllabus
Introduction
Stochastic convex optimization
Private empirical risk minimization
Private stochastic convex optimization
Snowball SGT
Privacy amplification
Reducing sensitivity
In iterative localization
Strongly convex
Summary
Taught by
Association for Computing Machinery (ACM)
Related Courses
-
Differentially Private Online-to-Batch Conversion for Stochastic Optimization
-
Private Convex Optimization via Exponential Mechanism - Differential Privacy for Machine Learning
-
Towards Practical Differentially Private Convex Optimization
-
Interaction Is Necessary for Distributed Learning with Privacy or Communication Constraints
-
Private Stochastic Optimization with Large Worst-Case Lipschitz Parameter
-
Large Scale Machine Learning and Convex Optimization - Lecture 3
