LGOCMLApr 16, 2024

Differentially Private Optimization with Sparse Gradients

arXiv:2404.10881v25 citationsh-index: 31NIPS
Originality Incremental advance
AI Analysis

This work addresses privacy-preserving optimization for large embedding models, offering incremental advances in efficiency and sparsity-based rates.

The paper tackles differentially private optimization for models with sparse gradients, achieving near-optimal bounds for mean estimation and dimension-independent rates for stochastic convex optimization, with improvements in high-dimensional regimes.

Motivated by applications of large embedding models, we study differentially private (DP) optimization problems under sparsity of individual gradients. We start with new near-optimal bounds for the classic mean estimation problem but with sparse data, improving upon existing algorithms particularly for the high-dimensional regime. Building on this, we obtain pure- and approximate-DP algorithms with almost optimal rates for stochastic convex optimization with sparse gradients; the former represents the first nearly dimension-independent rates for this problem. Finally, we study the approximation of stationary points for the empirical loss in approximate-DP optimization and obtain rates that depend on sparsity instead of dimension, modulo polylogarithmic factors.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes