Saved in:
Bibliographic Details
Main Authors: Kalinin, Nikita P., Upadhyay, Jalaj, Lampert, Christoph H.
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2502.06597
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910985153937408
author Kalinin, Nikita P.
Upadhyay, Jalaj
Lampert, Christoph H.
author_facet Kalinin, Nikita P.
Upadhyay, Jalaj
Lampert, Christoph H.
contents We propose Joint Moment Estimation (JME), a method for continually and privately estimating both the first and second moments of data with reduced noise compared to naive approaches. JME uses the matrix mechanism and a joint sensitivity analysis to allow the second moment estimation with no additional privacy cost, thereby improving accuracy while maintaining privacy. We demonstrate JME's effectiveness in two applications: estimating the running mean and covariance matrix for Gaussian density estimation, and model training with DP-Adam on CIFAR-10.
format Preprint
id arxiv_https___arxiv_org_abs_2502_06597
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Continual Release Moment Estimation with Differential Privacy
Kalinin, Nikita P.
Upadhyay, Jalaj
Lampert, Christoph H.
Machine Learning
We propose Joint Moment Estimation (JME), a method for continually and privately estimating both the first and second moments of data with reduced noise compared to naive approaches. JME uses the matrix mechanism and a joint sensitivity analysis to allow the second moment estimation with no additional privacy cost, thereby improving accuracy while maintaining privacy. We demonstrate JME's effectiveness in two applications: estimating the running mean and covariance matrix for Gaussian density estimation, and model training with DP-Adam on CIFAR-10.
title Continual Release Moment Estimation with Differential Privacy
topic Machine Learning
url https://arxiv.org/abs/2502.06597