Differentially Private Correlation Clustering

Abstract

Correlation clustering is a widely used technique in unsupervised machine learning. Motivated by applications where individual privacy is a concern, we initiate the study of differentially private correlation clustering. We propose an algorithm that achieves subquadratic additive error compared to the optimal cost. In contrast, straightforward adaptations of existing non-private algorithms all lead to a trivial quadratic error. Finally, we give a lower bound showing that any pure differentially private algorithm for correlation clustering requires additive error $\Omega$(n).

Cite

Text

Bun et al. "Differentially Private Correlation Clustering." International Conference on Machine Learning, 2021.

Markdown

[Bun et al. "Differentially Private Correlation Clustering." International Conference on Machine Learning, 2021.](https://mlanthology.org/icml/2021/bun2021icml-differentially/)

BibTeX

@inproceedings{bun2021icml-differentially,
  title     = {{Differentially Private Correlation Clustering}},
  author    = {Bun, Mark and Elias, Marek and Kulkarni, Janardhan},
  booktitle = {International Conference on Machine Learning},
  year      = {2021},
  pages     = {1136-1146},
  volume    = {139},
  url       = {https://mlanthology.org/icml/2021/bun2021icml-differentially/}
}