I want to show that for any vector $a\in \mathbb{R}^n$ with $\sum\limits_{i=1}^n a_i = 0$ and $m=\textrm{max}_i |a_i|$, it holds $$\frac{1}{n} \sum\limits_{i=1}^n e^{a_i} \leq \cosh (m)\, .$$ I really appreciate any ideas.


Let $n$ is even and $\max\limits_{i}a_i=m$.

Since $e^x$ is a convex function, by Karamata we obtain: $$\frac{1}{n}\sum_{i=1}^ne^{a_i}\leq\frac{1}{n}\left(\frac{n}{2}e^{-m}+\frac{n}{2}e^m\right)=\frac{e^m+e^{-m}}{2}=\cosh{m}.$$ The case $n$ is odd for you.

  • $\begingroup$ Thank you very much, i did not know Karamata's inequality and needed a moment to understand what majorization means. $\endgroup$ – Philmore May 26 '17 at 16:09
  • $\begingroup$ @Philmore I am ready to explain. By the way, I think it's better to wright $\max\limits_{i}|a_i|=m$. I am ready to fix my post if you want. $\endgroup$ – Michael Rozenberg May 26 '17 at 16:14
  • $\begingroup$ @Michael_Rozenberg Your post is great. Obviously one does not have to look at two different cases by looking at the double-sized vector with two copies of every element of $a$. I would have had a really hard time proving above inequality without your suggestion of using Karamata. $\endgroup$ – Philmore May 26 '17 at 20:05

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.