Private Stochastic Convex Optimization: Efficient Algorithms for Non-smooth Objectives
This work addresses privacy-preserving optimization for machine learning, offering efficient algorithms for non-smooth problems, though it appears incremental as it builds on existing noisy descent methods.
The authors tackled private stochastic convex optimization for non-smooth objectives by proposing a noisy mirror descent algorithm, achieving optimal rates in statistical complexity and query efficiency when privacy scales inversely with sample size.
In this paper, we revisit the problem of private stochastic convex optimization. We propose an algorithm based on noisy mirror descent, which achieves optimal rates both in terms of statistical complexity and number of queries to a first-order stochastic oracle in the regime when the privacy parameter is inversely proportional to the number of samples.