8
$\begingroup$

$\newcommand{\KL}{\operatorname{KL}}$Let $X$ be a Polish metric space and $P(X)$ the space of probability measures on $X$. Given $\mu, \nu\in P(X)$, recall that $$\KL(\mu\parallel\nu) = \begin{cases}\mathbb E_\mu[\log\tfrac{d\mu}{d\nu}]&\text{if $\mu\ll\nu$;}\\+\infty&\text{otherwise.}\end{cases}$$ I know that both $\mu\mapsto \KL(\mu\parallel\nu)$ and $\nu\mapsto \KL(\mu\parallel\nu)$ are convex maps $P(X)\to\mathbb [0,+\infty]$. Are $$\mu\mapsto \sqrt{\KL(\mu\parallel\nu)}$$ and $$\nu\mapsto \sqrt{\KL(\mu\parallel\nu)}$$ convex as well?

If this is not true in general, does it exists a measure $\mu\in P(X)$ such that $\nu\mapsto \sqrt{\KL(\mu\parallel\nu)}$ is a convex map $P(X)\to \mathbb [0,+\infty]$? Or a measure $\nu$ such that $\mu\mapsto \sqrt{\KL(\mu\parallel\nu)}$ is convex?

$\endgroup$
17
  • 3
    $\begingroup$ I am finally fully convinced that Civilized people need to send missionaries to that exotic tribe whose members are called Mathematicians, to acquaint them with spelling, etc. MathJax evolved from LaTeX and that evolved from TeX, and TeX was the invention of Donald Knuth. That he designed it as he did for the reasons for which he did is among the things that prove that he is a genius. And he was quite explicit in his book about this that uppermost in his mind was the purpose of making it easy to be fastidious about certain things, among them the reasons for my edits to this question. $\qquad$ $\endgroup$ Commented Feb 20, 2022 at 18:25
  • 1
    $\begingroup$ Note that $\mu\|\nu,$ coded as \mu\|\nu, looks different from $\mu\parallel\nu,$ coded as \mu\parallel\nu. I changed the former to the latter. Daily I see things typed by mathematicians using LaTeX that have convinced me that few among them even suspect that some sober-minded people think that things like this matter, and that there are good and substantial reasons for that. In particular, people who don't notice these differences are affected by them. $\endgroup$ Commented Feb 20, 2022 at 18:33
  • 2
    $\begingroup$ @MichaelHardy : Is there a reason to prefer \mu\parallel\nu to \mu\|\nu? To me, actually the latter looks better (even though I accepted your edits). $\endgroup$ Commented Feb 20, 2022 at 19:03
  • 2
    $\begingroup$ @MichaelHardy In general I am fastidious about $\rm\TeX$ but I agree with Iosif Pinelis that in this case, the parallel lines aren't really functioning as a binary relation. Kullback and Leibler originally used a colon rather than parallel lines. This was long before do-it-yourself typesetting, and extra space was not inserted around the colons. So for this particular case, I don't think there's clear right or wrong about how much space to insert. $\endgroup$ Commented Feb 21, 2022 at 14:52
  • 2
    $\begingroup$ @MichaelHardy : After all this discussion, since (i) $\mu$ and $\nu$ are, not operands, but the arguments of the explicitly given function $\text{KL}$ and (ii) the spaces in $\mu\parallel\nu$ look excessive to me, I have now replaced the instances of $\mu\parallel\nu$ by $\mu\,\|\,\nu$ (an intermediate version between $\mu\parallel\nu$ and $\mu\|\nu$), now using the code \,\|\,. Thank you and Timothy Chow for the editing and discussion. $\endgroup$ Commented Feb 21, 2022 at 15:23

1 Answer 1

10
$\begingroup$

$\newcommand\de\delta\newcommand{\KL}{\operatorname{KL}}\newcommand{\p}{\,\|\,}$The maps $$\mu\mapsto\sqrt{\KL(\mu\p\nu)}$$ and $$\nu\mapsto\sqrt{\KL(\mu\p\nu)}$$ are not convex in general.

Indeed, let $\mu_p:=p\de_0+(1-p)\de_1$, where $p\in(0,1)$ and $\de_a$ is the Dirac measure supported on $\{a\}$.

Then the second partial derivative with respect to $p$ of $\sqrt{\mathrm{KL}(\mu_p\p\mu_r)}$ at $(p,r)=(1/10,1/11)$ is $-7.17\ldots<0$. So, $\sqrt{\mathrm{KL}(\mu\p\mu_r)}$ is not convex in $\mu$.

Also, the second partial derivative with respect to $r$ of $\sqrt{\mathrm{KL}(\mu_p\p\mu_r)}$ at $(p,r)=(1/10,1/9)$ is $-11.50\ldots<0$. So, $\sqrt{\mathrm{KL}(\mu_p\p\nu)}$ is not convex in $\nu$.


You also asked: "If this is not true in general, does it exists a measure $\mu\in P(X)$ such that $\nu\mapsto \sqrt{\mathrm{KL}(\mu\p\nu)}$ is a convex map $P(X)\to \mathbb [0,+\infty]$? Or a measure $\nu$ such that $\mu\mapsto \sqrt{\mathrm{KL}(\mu\p\nu)}$ is convex?"

The answer to each of these two questions is yes, at least when $X=\{0,1\}$, say.

For $p\in(0,1)$, let
\begin{equation} F(p):=\sqrt{\mathrm{KL}(\mu_p\p\mu_{1/2})}, \end{equation} \begin{equation} f(p):=F''(p)4 \mathrm{KL}(\mu_p\p\mu_{1/2})^{3/2}, \end{equation} \begin{equation} f_1(p):=f'(p)(1-p)^2 p^2. \end{equation} Then $f_1(1/2)=f'_1(1/2)=f''_1(1/2)=0$ and \begin{equation} f'''_1(p)=\frac{2+4 p(1-p)}{(1-p)^2 p^2}>0. \end{equation} It follows that $F''(p)\ge0$, so that $\sqrt{\mathrm{KL}(\mu\p\mu_{1/2})}$ is convex in $\mu$.

For $r\in(0,1)$, let
\begin{equation} G(r):=\sqrt{\mathrm{KL}(\mu_{1/2}\p\mu_r)}, \end{equation} and \begin{equation} g(r):=G''(r)4\mathrm{KL}(\mu_{1/2}\p\mu_r)^{3/2}. \end{equation} Then $g(1/2)=g'(1/2)=g''(1/2)=g'''(1/2)=0$ and \begin{equation} g''''(1/2+h)\frac{(1 - 4 h^2)^4}{16}=9- 16 h^4 + 156 h^2 + 64 h^6 \\ >9- 1 + 156 h^2 + 64 h^6>0 \end{equation} if $|h|<1/2$. It follows that $G''(r)\ge0$, so that $\sqrt{\KL(\mu_{1/2}\p\nu)}$ is convex in $\nu$.


Remark 1: The existence of a probability measure $\nu\in P(X)$ such that $\sqrt{\KL(\mu\p\nu)}$ is convex in $\nu$ holds for any Polish space $X$. Indeed, the case when $X$ is a singleton set is trivial. Otherwise, take any distinct points $x$ and $y$ in $X$ and let $\nu:=\frac12\,\de_x+\frac12\,\de_y$. Then the condition $\mu\ll\nu$ implies $\mu=p\,\de_x+(1-p)\,\de_y$ for some $p\in[0,1]$, so that here $\sqrt{\KL(\mu\p\nu)}=F(p)$.

Remark 2: Let us say that a probability measure $\mu\in P(X)$ is good if $\sqrt{\KL(\mu\p\nu)}$ is convex in $\nu$. Then it is easy to see that the Dirac measure $\de_x$ is good for any $x\in X$. It also follows from above that $\mu:=\frac12\,\de_x+\frac12\,\de_y$ is good if $X=\{x,y\}$ for some $x,y$.

Finally, using reasoning similar to that above, one can show that, in the case when the cardinality of $X$ is $\ge3$, the only good probability measures $\mu\in P(X)$ are the Dirac measures. Since this answer has already grown rather long, I will leave the latter assertion as an exercise to interested readers.

Now the answer is quite complete.

$\endgroup$
0

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .