How to prove $k(x_i,x_j)=e^{-(LR(x_i-x_j))^TLR(x_i-x_j)}$ is a valid kernel function or positive semi definite?

$x=(\mu,\lambda)^T$ and R is a 2x2 rotation matrix, L is a 2x2 diagonal scaling matrix with positive entries. Any idea is appreciated.


One very routine proof is to employ Schoenberg's interpolation theorem (see chapter 15 of A Course in Approximation Theory by Cheney and Light or sec. 2.5 of this book chapter, for instances).

The radial basis function $\phi(y) = \exp(-y)$ is completely monotone on $[0,\infty)$. By Schoenberg's interpolation theorem , $\phi(\|\cdot\|^2)$ is strictly positive definite. Therefore $f(u,v)=\phi(\|u-v\|^2)=e^{-\|u-v\|^2}$ is a kernel. As $LR$ is invertible, it follows that the matrix $\left(k(x_i,x_j)\right)_{i,j=1,2,\ldots,n}=\left(f(LRx_i,LRx_j)\right)_{i,j=1,2,\ldots,n}$ is positive definite if all the $x_i$s are distinct, or positive semidefinite otherwise.

  • $\begingroup$ How to derive this step $(k(x_i,x_j))_{i,j=1,2,…,n}=(f(LRx_i,LRx_j))_{i,j=1,2,…,n}$ ? $\endgroup$ – user40577 Feb 27 '14 at 3:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.