Answers are grounded in this paper and cite their sources.
The main result is that the eigenvalues of any continuously differentiable positive definite kernel are o(1 / n² ). The method of proof is to approximate the kernel by kernels of finite rank in such a way that the difference is positive definite. The tail of the eigenvalue series can then be related to the trace integral of the difference by means of Mercer’s theorem and a trace norm version of the Weyl–Courant minimax principle. It is conjectured that, for p times continuously differentiable positive definite kernels, the eigenvalues are o(1 / np + 1 ).
No takes yet. Share an insight, caveat, or question.
J. B. Reade (1983) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: