Not All Learnable Distribution Classes are Privately Learnable

We give an example of a class of distributions that is learnable in total variation distance with a finite number of samples, but not learnable under $(\varepsilon, \delta)$-differential privacy. This refutes a conjecture of Ashtiani.

Gespeichert in:

Bibliographische Detailangaben
Hauptverfasser:	Bun, Mark, Kamath, Gautam, Mouzakis, Argyris, Singhal, Vikrant
Format:	Artikel
Sprache:	eng
Schlagworte:	Computer Science - Cryptography and Security Computer Science - Data Structures and Algorithms Statistics - Machine Learning
Online-Zugang:	Volltext bestellen
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!

Beschreibung
Zusammenfassung:	We give an example of a class of distributions that is learnable in total variation distance with a finite number of samples, but not learnable under $(\varepsilon, \delta)$-differential privacy. This refutes a conjecture of Ashtiani.
DOI:	10.48550/arxiv.2402.00267