[edit]

# Not All Learnable Distribution Classes are Privately Learnable

*Proceedings of The 35th International Conference on Algorithmic Learning Theory*, PMLR 237:390-401, 2024.

#### Abstract

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.