An exact toy example: successive Mercer truncations approach a slice of the Brownian kernel K(x, y) = min(x, y). Dashed curves show the exact slice. The paper studies how smoothness and spectral decay control uniform approximation error. Research source
About this paper
Quantifies how quickly truncated Mercer expansions converge uniformly to a positive definite kernel.
The convergence estimates depend on eigenvalue tails and kernel smoothness. The results also give spectral information about integral operators whose roots and other powers have continuous kernels.
Citation
BibTeX
@article{takhanov2023merceruniformconvergence,
title = {On the speed of uniform convergence in Mercer’s theorem},
author = {Takhanov, Rustem},
year = {2023},
journal = {Journal of Mathematical Analysis and Applications},
eprint = {2205.00487},
archivePrefix = {arXiv},
url = {https://arxiv.org/pdf/2205.00487}
}