$d(x,y)$ in terms of reproducing kernels for a H separating points

109 Views Asked by At

I'm working through an introduction to the Theory of Reproducing Kernel Hilbert Spaces by V.I. Paulsen and M. Raghupathi.

Exercise 1.1 states, that let $\mathcal{H}$ be the reproducing kernel hilbert space of $X$, $d(x,y,):=\sup\{|f(x)-f(y)|: f\in\mathcal{H}, \|f\|\le 1\}$ is a metric if and only if $\mathcal{H}$ separates points.

I was able to prove this statement. However, they go on and say give a formula for $d(x,y)$ in terms of the reproducing kernel. What I achieved so far

$$|f(x)-f(y)|=\|\langle f,k_x\rangle-\langle f,k_y\rangle\| \le\|f\|\|k_x-k_y\|\le\|k_x-k_y\|$$

i.e. $d(x,y)\le\|k_x-k_y\|$. Now clearly would be nice to show that this bonudary is sharp and I guess if this is the case we would need the separation property. Can someone give me a hint, please?

1

There are 1 best solutions below

1
On BEST ANSWER

As you note you have that $$|f(x)-f(y)|=| \langle f, k_x-k_y\rangle| ≤ \|k_x-k_y\|$$

Giving that $\sup_{\|f\|≤1}|f(x)-f(y)|≤\|k_x-k_y\|$. Now look at $f= \frac{k_x-k_y}{\|k_x-k_y\|}$, which is well defined (since $\mathcal H$ separates points) and norm one. You have that

$$\langle\frac{k_x-k_y}{\|k_x-k_y\|}, k_x-k_y\rangle = \frac{\|k_x-k_y\|^2}{\|k_x-k_y\|} =\|k_x-k_y\|$$ hence $\sup_{\|f\|≤1} |f(x)-f(y)|≥\|k_x-k_y\|$.