CS-233: SVM Gaussian Kernel

Hello,

In the theory slides "High-Dimensional Features and KernelsFile" at page 40 it is said that the gaussian Kernel is of the form exp(- (||x-x'||^2 ) / (sigma^2)) . I wonder if it is a typo, or if it was a choice to omit the 1/2 factor in the exponent, since Gaussian models usually are of the form exp(- (||x-x'||^2 ) / (2*sigma^2)).

Thank's for considering this.

Yours sincerely,

Lucas

Re: SVM Gaussian Kernel

by Vidit Vidit - Tuesday, 18 June 2019, 2:19 PM

Hello,
Yes, that is a mistake.
$k(x,x^\prime)=\exp(\frac{-\||x-x^\prime\||^2}{2\sigma^2})$ is the correct formulation.

Cheers