Reproducing kernel Hilbert space: Difference between revisions

Revision as of 07:26, 29 April 2025

Revision as of 04:51, 8 May 2025

Line 140:

Mercer's theorem states that the spectral decomposition of the integral operator <math>T_K</math> of <math>K</math> yields a series representation of <math>K</math> in terms of the eigenvalues and eigenfunctions of <math> T_K </math>. This then implies that <math>K</math> is a reproducing kernel so that the corresponding RKHS can be defined in terms of these eigenvalues and eigenfunctions. We provide the details below.

Under these assumptions <math>T_K</math> is a compact, continuous, self-adjoint, and positive operator. The [[spectral theorem]] for self-adjoint operators implies that there is an at most countable decreasing sequence <math>(\sigma_i)_i_{i \geq 0} </math> such that <math display="inline">\lim_{i \to \infty}\sigma_i = 0</math> and

<math>T_K\varphi_i(x) = \sigma_i\varphi_i(x)</math>, where the <math>\{\varphi_i\}</math> form an orthonormal basis of <math>L_2(X)</math>. By the positivity of <math>T_K, \sigma_i > 0</math> for all <math>i.</math> One can also show that <math>T_K </math> maps continuously into the space of continuous functions <math>C(X)</math> and therefore we may choose continuous functions as the eigenvectors, that is, <math>\varphi_i \in C(X)</math> for all <math>i.</math> Then by Mercer's theorem <math> K </math> may be written in terms of the eigenvalues and continuous eigenfunctions as