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

Let $(g_k)$ be a sequence of independent standard gaussians variables on a fixed probability space $\Omega$. Let $(\epsilon_k)$ be a sequence of independent rademacher variables.

What is the best constant in the following inequality: $$ \vert\vert\sum_{k}\epsilon_k \otimes x_k \vert\vert_{L^2(\Omega,E)} \leq K \vert\vert\sum_{k}g_k\otimes x_k \vert\vert_{L^2(\Omega,E)}\ \ ? $$ for any Banach space $E$, any $x_1,\ldots, x_n \in E$. I know that the best constant is $\leq \sqrt{\frac{\pi}{2}}$ (see Diestel,Jarchow,Tonge "Absolutely summing operators" page 239).

share|cite|improve this question
How do you define "standard gaussians variables" on a "fixed probability space $\Omega$"? – Ricky Demer May 13 '11 at 20:02
Ricky: I don't think that is going to be an issue/problem here. There are standard technical theorems in the literature ensuring that one can create a suitable $\Omega$ and $\sigma$-algebra so that such random variables can be defined. How one constructs them does not (IIRC) affect what $K$ can be chosen – Yemon Choi May 13 '11 at 20:20
up vote 9 down vote accepted

$\sqrt{\pi\over 2}$, the reciprocal of the $L_1$ norm of a standard gaussian, is the best constant. Let $x_k$ be the kth unit basis vector in $\ell_1$ and let the sum go from $1$ to $N$. The square of the left hand side is $N^2$ and the square of the right hand side is $N+N\sqrt{2\over \pi}(N-1)\sqrt{2\over \pi}$ (multiplied by $K^2$).

share|cite|improve this answer

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.