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.

翻译：我们给出一个分布类的例子，该分布类在总变差距离下可用有限数量的样本进行学习，但在$(\varepsilon, \delta)$-差分隐私下却不可学习。这一结果反驳了Ashtiani的一个猜想。