1: \begin{abstract}

2:    This paper characterizes the maximum mean discrepancies (MMD) that metrize

3:    the weak convergence of probability measures for a wide class of kernels.

4:    More precisely,

5:    % Theorem~12 of Simon-Gabriel \& Schölkopf (JMLR, 2018, \cite{simon18kde}) seemed to close a k40-year-old quest to characterize maximum mean discrepancies (MMD) that metrize the weak convergence of probability measures.

6:    % We prove, however, that the theorem is incorrect and provide a correction.

7:    we prove that, on a

8:    locally compact, non-compact, Hausdorff space, the MMD of a bounded continuous Borel measurable kernel $\k$, whose RKHS-functions vanish at infinity (i.e.,\ $\Hk

9:    \subset  \C{}{0}$), metrizes the weak convergence of probability measures if

10:    and only if $\k$ is continuous and \ispd over all signed, finite, regular

11:    Borel measures. 

12:    %The additional assumption $\Hk \subset \C{}{0}$ cannot be dropped: 

13:    We also correct a prior result of 

14:    %Simon-Gabriel \& Schölkopf (JMLR, 2018,

15:    \citet[JMLR][Thm.~12]{simon18kde}

16:    %, Thm.~12) 

17:    by showing that

18:    % this result does not hold for all bounded continuous $k$ as there

19:    there exist both bounded continuous \ispd kernels that do not metrize weak convergence and bounded continuous non-\ispd kernels that do metrize it.

20: \end{abstract}

21: