tensorflow_privacy/research/dp_newton

Project Title

Faster Differentially Private Convex Optimization via Second-Order Methods https://arxiv.org/abs/2112.03570
by Arun Ganesh, Mahdi Haghifam, Thomas Steinke, Abhradeep Thakurta.

Description

Implementation of the optimizatoin algorithms proposed in https://arxiv.org/abs/2112.03570

Getting Started

You will need to install fairly standard dependencies

run 'run_privacy_utility' to compare the convergence speed and excess loss of different algorithms.

Citation

You can cite this paper with

@article{ganesh2023faster,
  title={Faster Differentially Private Convex Optimization
    via Second-Order Methods},
  author={Ganesh, Arun and Haghifam, Mahdi and Steinke, Thomas
    and Thakurta, Abhradeep},
  journal={arXiv preprint arXiv:2305.13209},
  year={2023}
}