If a function is positive definite,then matrix is P.S.D. x1,,,,xn⊂X=>K0(xi,xj)=g(xi)g(xj) =>k0=[g(x1),..,g(xn)]′∗[g(x1),...,g(xn)]
Thm:
Let F be a probalility measure on the half low Pat such that 0<∫0∞sdF(s)<∞ and l(F,u)=∫0∞exp(−tsϕ)dF is P.D for all t>0; example: polynomial kernel. RBF Gauss kernel. two advantages:1.lowdimension−>∞dimension 2.normalize.
Levy distribution
(B/2∗pi)1/2exp(sqrt(2B))∣f(s)=sqrt(t/2∗pi)u−3/2exp(−t/2u)du if ϕ(x)=K+1/2 K21∗K21=KT
Thm
let kX∗X−>R be a P.D kernel then exists a HILBERT space H and from x->H such that ϕ(H) ∀x,y⊂x,K(x,y)=<ϕ(x),ϕ(y)> three kernels.