Total Variation Distance for Product Distributions is $\#\mathsf{P}$-Complete

Abstract

We show that computing the total variation distance between two product distributions is $\#\mathsf{P}$-complete. This is in stark contrast with other distance measures such as Kullback-Leibler, Chi-square, and Hellinger, which tensorize over the marginals leading to efficient algorithms.

5 pages. An extended version of this paper appeared in the proceedings of IJCAI 2023, under the title "On approximating total variation distance" (see https://www.ijcai.org/proceedings/2023/387 and arXiv:2206.07209)