MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

What is the easiest (preferably without calculations) way to see that the mean value of $\max(x_1,x_2,\dots,x_n)$ on the sphere $\mathbb{S}^{d-1}= \{ (x_1,\dots,x_n):\ x_1^2+\dots+x_n^2=1 \}$ behaves like $\sqrt{\log(n)/n}$, or at least that is is much more then $1/\sqrt{n}$ for large $n$? The same (and less or more a priori equivalent) question concerns the standard Gaussian measure and expectation of $\infty$-norm w.r.t. it.

The proofs I know (for example, the one which V. Milman attributes to Figiel) use too many integrals.

And by the way, how to put {,} in math here? \ { does not work for me

share|cite|improve this question
\lbrace and \rbrace. – Gerry Myerson Nov 9 '10 at 12:01
Or \\{ and \\}. – Andrey Rekalo Nov 9 '10 at 12:02
Oh, thanks! $\ \ $ – Fedor Petrov Nov 9 '10 at 12:50
doesn't the fact that on your hypersphere ($d$ should be $n$ in your question), $\|x\|_\infty \ge 1/\|x\|_1 \ge 1/\sqrt{n}$ help? – Suvrit Nov 10 '10 at 15:04
up vote 15 down vote accepted

For some positive $c$ bounded away from zero, the probability that a standard gaussian variable is larger than $c\sqrt{\log n}$ is $1/n$. It follows that the probability that at least one variable out of $n$ independent standard gaussians is larger than $c\sqrt{\log n}$ is $1-(1-1/n)^n$ which tends to $1-1/e$. From that one gets that the expectation of the maximum of $n$ standard gaussians is at least (basically) $(1-1/e)c\sqrt{\log n}$. As you said the question about variables on the sphere is equivalent to that.

share|cite|improve this answer
Thanks, that's clear indeed. – Fedor Petrov Nov 9 '10 at 12:53

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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