activity
20242026
most citedBack to Square Roots: An Optimal Bound on the Matrix Factorization Error for Multi-Epoch Differentially Private SGD

1 citations · 1 across the 1 of their papers we have counts for

collaborators

7 papers

cs.CR20261 cited

Back to Square Roots: An Optimal Bound on the Matrix Factorization Error for Multi-Epoch Differentially Private SGD

Nikita P. Kalinin, Ryan McKenna, Jalaj Upadhyay +1

Matrix factorization mechanisms for differentially private training have emerged as a promising approach to improve model utility under privacy constraints. In practical settings,…

cs.LG2025

Continual Release Moment Estimation with Differential Privacy

Nikita P. Kalinin, Jalaj Upadhyay, Christoph H. Lampert

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 approache…

cs.DS2025

On the Price of Differential Privacy for Hierarchical Clustering

Chengyuan Deng, Jie Gao, Jalaj Upadhyay +2

Hierarchical clustering is a fundamental unsupervised machine learning task with the aim of organizing data into a hierarchy of clusters. Many applications of hierarchical clusteri…

cs.DS2025

A Generalized Binary Tree Mechanism for Differentially Private Approximation of All-Pair Distances

Michael Dinitz, Chenglin Fan, Jingcheng Liu +2

We study the problem of approximating all-pair distances in a weighted undirected graph with differential privacy, introduced by Sealfon [Sea16]. Given a publicly known undirected…

cs.CR2024

Almost linear time differentially private release of synthetic graphs

Jingcheng Liu, Jalaj Upadhyay, Zongrui Zou

In this paper, we give an almost linear time and space algorithms to sample from an exponential mechanism with an -score function defined over an exponentially large non-co…

cs.CR2024

Optimality of Matrix Mechanism on -metric

Jingcheng Liu, Jalaj Upadhyay, Zongrui Zou

In this paper, we introduce the -error metric (for ) when answering linear queries under the constraint of differential privacy. We characterize such an error u…