1
$\begingroup$

I would like to minimize the following function:

$f(x)=log(e^{-x_1}+..+e^{-x_n})$

Subject to:

$\sum_{i=1}^{n}{x_i}=1$

$0 \leq x_i \leq 1$

So far I have discovered the following: If all the $x_i$'s are equal, $f(x)=max(x_i)+log(n)$, but I have not been able to find a conclusive answer regarding what would be the value of the $x_i$'s for the minimum value of the function.

|cite|improve this question
$\endgroup$
  • $\begingroup$ Min occur at $x_i = \frac{1}{n}$. $\endgroup$ – Mhenni Benghorbal May 4 '14 at 20:58
1
$\begingroup$

Let $y_i = e^{-x_i}$, then $0 < y_i \leq 1$, and $x_i = -lny_i$, and then $-lny_1 - lny_2 -...-lny_n = 1 \to ln(y_1y_2....y_n) = -1 \to y_1y_2...y_n = \dfrac{1}{e}$. So: by AM-GM inequality: $log(y_1 + y_2 +...+ y_n) \geq log(n\cdot (y_1y_2...y_n)^{\frac{1}{n}}) = log(n\cdot e^{-1/n}) = logn - \dfrac{1}{n}$.

So: $f_{min} = logn - \dfrac{1}{n}$

|cite|improve this answer
$\endgroup$
  • $\begingroup$ Is this a lower bound on the minimum? Or is this the minimum? From the proof it seems to be a lower bound on the minimum. $\endgroup$ – ASDF May 5 '14 at 16:59
  • $\begingroup$ @ASDF: it is the minimum. $\endgroup$ – DeepSea May 5 '14 at 19:33
1
$\begingroup$

log is increasing, so the minimum occurs at the same place as the minimum of $\sum e^{-x_i}$; and $x\mapsto e^{-x}$ is convex, so $\sum e^{-x_i} \ge ne^{-\frac1n\sum x_i} = ne^{-1/n}$, with equality when all $x_i$ are equal.

|cite|improve this answer
$\endgroup$

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.