5
$\begingroup$

Let $U$ be an open convex set in a locally convex space $X$, and let $f : U \to [0,1]$ be a log-concave function on $U$ (i.e., bounded and real-valued). Under what conditions does $f$ have a continuous extension to the closure $\overline U$?

$\endgroup$
6
  • $\begingroup$ My simple-minded reaction to this is to study $F = e^{-f}$, which is convex and bounded (if you assume the image of $f$ is in $[0,1]$). The properties of convex functions are rather well known. $\endgroup$
    – Deane Yang
    Nov 24, 2010 at 18:55
  • $\begingroup$ Thanks for the thought, Deane. Here it's that $-\log f$ is convex, not $\mathrm e^{-f}$. Originally, I had phrased the question in terms of convex functions instead. Since $-\log 0 = \infty$, though, I figured the question would be clearer if I just asked the log-concave version. You're right: this question translates to one about convex functions, which I also do not know the answer to. $\endgroup$ Nov 24, 2010 at 20:32
  • $\begingroup$ Define $f(0) = 0$ and $f(1) = \mathrm e^{-1}$. I don't understand the point of your example; perhaps you misunderstood my question? $\endgroup$ Nov 24, 2010 at 22:17
  • $\begingroup$ Tom, sorry for the dyslexic comment. $\endgroup$
    – Deane Yang
    Nov 24, 2010 at 22:49
  • $\begingroup$ No problem, Deane. It was a good suggestion in intent, and will perhaps help somebody answer the question. $\endgroup$ Nov 24, 2010 at 22:54

1 Answer 1

3
$\begingroup$

I do not see how $\log$-concavity should imply any form of continuity. For instance, if $\|\cdot\|$ is any semi-norm on the locally convex space $X$ then $f(x) = e^{-\|x\|}$ will be bounded and $\log$-concave but it will only be continuous if the semi-norm is continuous.

What you really want in order to be able to extend $f$ to the boundary of $U$ is uniform continuity and for that the fact that $f$ is $\log$-concave may help but is certainly not sufficient.

$\endgroup$
1
  • $\begingroup$ Yes, certainly if $f$ is not even continuous on $U$, then it has no continuous extension to the closure... Perhaps the wording of the problem needs to be changed. $\endgroup$ Nov 25, 2010 at 14:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.