Arun Ganesh
I am a Research Scientist at Google Research based in Seattle. I am interested in differential privacy, especially private optimization and privacy accounting/amplification.
Before joining Google, I got my PhD at UC Berkeley, where I was in the Theory Group advised by Satish Rao.
Email: <firstname><lastname>@google.com
Publications
(All co-author lists are ordered alphabetically here; this may differ from the ordering on arXiv or in the conference version)
Privacy Amplification for Matrix Mechanisms - with Christopher A. Choquette-Choo, Thomas Steinke, Abhradeep Thakurta
arXiv, ICLR 2024
Correlated Noise Provably Beats Independent Noise for Differentially Private Learning - with Christopher A. Choquette-Choo, Krishnamurthy Dvijotham, Krishna Pillutla, Thomas Steinke, Abhradeep Thakurta
arXiv, ICLR 2024
(Amplified) Banded Matrix Factorization: A unified approach to private training - with Christopher A. Choquette-Choo, Ryan McKenna, Brendan McMahan, Keith Rush, Abhradeep Thakurta, Zheng Xu
arXiv, Neurips 2023
Faster Differentially Private Convex Optimization via Second-Order Methods - with Mahdi Haghifam, Thomas Steinke, Abhradeep Thakurta
arXiv, Neurips 2023
Private (Stochastic) Non-Convex Optimization Revisited: Second-Order Stationary Points and Excess Risks - with Daogao Liu, Sewoong Oh, Abhradeep Thakurta
arXiv, Neurips 2023
Universality of Langevin Diffusion for Private Optimization, with Applications to Sampling from Rashomon Sets - with Abhradeep Thakurta, Jalaj Upadhyay
arXiv, COLT 2023
Why is Public Pretraining Necessary for Private Model Training? - with Mahdi Haghifam, Milad Nasr, Sewoong Oh, Thomas Steinke, Om Thakkar, Abhradeep Thakurta, Lun Wang
arXiv, ICML 2023
How Compression and Approximation Affect Efficiency in String Distance Measures - with Andrea Lincoln, Barna Saha, Tomasz Kociumaka
arXiv, SODA 2022
Public Data-Assisted Mirror Descent for Private Model Training - with Ehsan Amid, Rajiv Mathews, Swaroop Ramaswamy, Shuang Song, Thomas Steinke, Vinith M. Suriyakumar, Abhradeep Thakurta, Om Thakkar
arXiv, ICML 2022
Universal Algorithms for Clustering Problems - with Bruce M. Maggs, Debmalya Panigrahi
arXiv, ICALP 2021
Privately Answering Counting Queries with Generalized Gaussian Mechanisms - with Jiazheng Zhao
arXiv, FORC 2021
Faster Differentially Private Samplers via Rényi Divergence Analysis of Discretized Langevin MCMC - with Kunal Talwar
arXiv, Neurips 2020
Near-Linear Time Edit Distance for Indel Channels - with Aaron Sy
arXiv, WABI 2020
Robust Algorithms for Steiner Tree and TSP - with Bruce M. Maggs, Debmalya Panigrahi
arXiv, ICALP 2020
Optimal Sequence Length Requirements for Phylogenetic Tree Reconstruction with Indels - with Qiuyi Zhang
arXiv, STOC 2019
Online Service with Delay - with Yossi Azar, Rong Ge, Debmalya Panigrahi
arXiv, STOC 2017
Manuscripts
Near Exact Privacy Amplification for Matrix Mechanisms - with Christopher A. Choquette-Choo, Saminul Haque, Thomas Steinke, Abhradeep Guha Thakurta
The Last Iterate Advantage: Empirical Auditing and Principled Heuristic Analysis of Differentially Private SGD - with Borja Balle, Christopher A. Choquette-Choo, Matthew Jagielski, Jamie Hayes, Milad Nasr, Adam Smith, Thomas Steinke, Andreas Terzis, Abhradeep Guha Thakurta
Optimal Rates for O(1)-Smooth DP-SCO with a Single Epoch and Large Batches - with Christopher A. Choquette-Choo, Abhradeep Thakurta
Fine-Tuning Large Language Models with User-Level Differential Privacy - with Zachary Charles, Ryan McKenna, Brendan McMahan, Nicole Mitchell, Krishna Pillutla, Keith Rush
Tight Group-Level DP Guarantees for DP-SGD with Sampling via Mixture of Gaussians Mechanisms
This work was partially subsumed by Fine-Tuning Large Language Models with User-Level Differential Privacy
Recycling Scraps: Improving Private Learning by Leveraging Intermediate Checkpoints - with Virat Shejwalkar, Rajiv Mathews, Om Thakkar, Abhradeep Thakurta