Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Let $y,z\in(0,1)^n$ satisfy $||y||_1 = ||z||_1=1$.

Then $$ \frac{||z||_3}{||z||_2} \le K_n ||z/y||_\infty \frac{||y||_3}{||y||_2} $$ where $z/y\in\mathbb R^n$ is the coordinate-wise quotient of $z$ and $y$ and $K_n>0$ is some constant depending only on $n$. What's the best $K_n$ that makes this true? What if $2,3$ are replaced by general $p,q$?

share|improve this question
    
It's easy to see that $1 \le K_n \le \sqrt{n}$. Do you really need the best constant, or just some improvement on this? –  Mark Meckes Mar 3 '11 at 14:59
    
For such a strange question you should give some motivation and say what you know. –  Bill Johnson Mar 3 '11 at 16:15
1  
My motivation is to obtain a Berry-Esseen type bound for the case where the coefficients are generated by a certain kind of partition. This probably isn't making a whole lot of sense; I'll write post a link to a short writeup soon. –  Aryeh Kontorovich Mar 3 '11 at 17:27
    
@Mark -- the most interesting question is whether the best $K_n$ is actually a constant (which I suspect is not the case). The next question is -- does it grow as $n$ to some powr or slower (logarithmically, say)? The $\sqrt n$ bound is unfortunately too crude for my needs. –  Aryeh Kontorovich Mar 3 '11 at 22:29
    
Here is a short writeup providing some motivation: dl.dropbox.com/u/3198145/norm-ineq.pdf –  Aryeh Kontorovich Mar 3 '11 at 23:14

Your Answer

 
discard

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

Browse other questions tagged or ask your own question.