Faster Differentially Private Top-$k$ Selection: A Joint Exponential Mechanism with Pruning
Cryptography and Security
2026-01-09 v1
Abstract
We study the differentially private top- selection problem, aiming to identify a sequence of items with approximately the highest scores from items. Recent work by Gillenwater et al. (ICML '22) employs a direct sampling approach from the vast collection of possible length- sequences, showing superior empirical accuracy compared to previous pure or approximate differentially private methods. Their algorithm has a time and space complexity of . In this paper, we present an improved algorithm with time and space complexity , where denotes the privacy parameter. Experimental results show that our algorithm runs orders of magnitude faster than their approach, while achieving similar empirical accuracy.
Cite
@article{arxiv.2411.09552,
title = {Faster Differentially Private Top-$k$ Selection: A Joint Exponential Mechanism with Pruning},
author = {Hao WU and Hanwen Zhang},
journal= {arXiv preprint arXiv:2411.09552},
year = {2026}
}
Comments
NeurIPS 2024