Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. 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

I'm having trouble digesting the proof of Theorem 9.6 in Milman and Schechtman's classic book "Asymptotic theory of finite dimensional normed spaces" (pg. 54). I'm new to functional analysis, so this is surely a simple misunderstanding on my part.

Theorem 9.6

Here, $M_{\|~\|}$ is denoting the average norm for points on $S^{n-1}$. The first line in the computation of $M_{\|\cdot\|}$ is confusing me:

$$M_{\|\cdot\|} \approx \left(\int_{S^{n-1}} \|\sum a_i e_i\|^2 d\mu(\bar{a})\right)^{1/2}$$

We're actually trying to show a lower bound here, so isn't the $\approx$ hiding a $1/\sqrt{n}$ factor? Isn't this bad for the rest of the argument? Earlier in the book, in the proof of Dvoretzky's Theorem, $M_{\|\cdot\|}$ is given as $\int_{S^{n-1}} \|\sum a_i e_i\| d\mu(\bar{a})$, which is what I'd have expected here also.

share|cite|improve this question
Are you confusing the median $M_{\|\ \|}$ with the mean/average $A_{\|\ \|}$? – user31373 Jul 14 '12 at 19:11
up vote 1 down vote accepted

By Remark 1 at the bottom of page 19, the median $M_{\|\ \|}$ is within an additive constant $C_p$ from the $L^p$ mean $\left(\int_{S^{n-1}}\|x\|^p\right)^{1/p}$, for any $1\le p<\infty$. They use this remark with $p=2$. John's ellipsoid theorem 3.3 is invoked here only to justify the application of 5.1 and the subsequent remark 1. They do not use $\|x\|\ge |x|/\sqrt{n}$ under any of the integral signs; as you said, this would destroy the argument. But the better bound $\|e_i\|\ge 1/4$ holds for half of basis vectors, and this is enough to make the cotype inequality effective.

share|cite|improve this answer
I'd indeed confused median with mean, and I didn't remember that remark. Thanks for your help! – arnab Jul 14 '12 at 21:06

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.