I asked the following question on stackexchange some time ago. Let $X,Y\in \mathbb{R}^d$ with $d\ge 1$. Also $X\sim N(0, \Sigma)$ and $Y$ (general r.v.) has zero mean and the same variance $\Sigma$. I'm looking for the following bound:

$$ \mathbb{E}\|X\|_{\infty}^r\le C_{d,r}\mathbb{E}\|Y\|_{\infty}^r $$

for some constant $C_{d,r}$ depending on the dimension of vectors $X$ and $Y$ ($d$) and $r\ge 2$. It's easy to see that we have

$$ \mathbb{E}\|X\|_{\infty}^r \le C_r d^{r/2}\mathbb{E}\|Y\|_{\infty}^r, $$

where $C_r$ is a constant depending on $r$. The latter follows from the fact that $\|x\|_{\infty}\le \|x\|_2\le \sqrt{d}\|x\|_{\infty}$ for $x\in \mathbb{R}^d$ and a known inequality regarding $\|\cdot\|_2$, i.e. $$ \mathbb{E}\|X\|_2^r\le C_r'\left(\mathbb{E}\|X\|_2^2\right)^{r/2}. $$

However, can we have $C_{d,r}$ of order $C_r \ln(d)$ or smaller?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Browse other questions tagged or ask your own question.