Houman Owhadi: On learning adapted kernels. One World Numerical Analysis Series.

Abstract: Most numerical approximation methods can be interpreted as kernel interpolation methods. The particular choice of the kernel can have a large impact on the accuracy/complexity of the method and this talk is an invitation to explore the problem of selecting/learning the underlying kernel.

Talk given on Nov 2nd, 2020 at the One World Numerical Analysis Series.

and

Covers material found in the following book
and the following papers