0
$\begingroup$

If the world is nice, let $f$ be a positive and strictly increasing differentiable function, i.e. $f>0$ and $f'>0$.

  1. Is the following statement true?

    There always exist $p\in\mathbb{R}$, such that $g = f^p$ is convex.

  2. How about we replace 'convex' with 'concave'?

  3. I came across this question in my study of submodularity, a notion of concavity in combinatorial optimization. If you are familiar with submodularity, is it true that for any monotonic function $f$, I can take some power of $f$ and make it submodular?

$\endgroup$
1
$\begingroup$

No, for both of them take $f(x) =\arctan(x) + \pi$,

And note that for a convex(concave) function $g$ defined on $R$ , we have $\limsup_{|x| \rightarrow \infty} |g(x)| = \infty $

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ That's a good counterexample. How about I change f'>0 to f' is bounded away from 0? $\endgroup$ – Xuezhou Zhang Jun 10 '17 at 5:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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