PDA

View Full Version : Hilbert schmidt norm


dirk_mec1
Dec3-08, 02:28 PM
1. The problem statement, all variables and given/known data

http://img523.imageshack.us/img523/4456/56166304yr3.png (http://imageshack.us)

2. Relevant equations
http://img356.imageshack.us/img356/2793/40249940is8.png (http://imageshack.us)


3. The attempt at a solution
I defined K:[a,b] --> [a,b] with k(s,t) = \frac{(t-s)^{n-1}}{(n-1)!}

I found for the norm:

\int_a^b \int_a^b \frac{(t-s)^{2n-2}}{(n-1)!^2}\ \mbox{d}s\ \mbox{d}t =0

Is this correct?

morphism
Dec3-08, 09:48 PM
If the norm of blah is zero, then blah is zero. Is blah zero in this case?

dirk_mec1
Dec4-08, 05:02 AM
If the norm of blah is zero, then blah is zero. Is blah zero in this case?

You're right something is wrong.

But is the integral set up with the correct boundaries?

dirk_mec1
Dec5-08, 03:32 AM
Ok presuming the boundaries are ok I end up with:

||A||_{HS} = \frac{2 (b-a)^n}{((n-1)!)^2 (2n-1)(2n)}

Is this correct?

morphism
Dec5-08, 10:36 AM
Did you remember to take the square root?

Pere Callahan
Dec5-08, 11:16 AM
You might want to include some characteristic function like \chi_{\{s\leq t\}} in your kernel function.