Relative concentration bounds for the kernel matrix spectrum

  • 2018-12-05 16:58:07
  • Ernesto Araya Valdivia
  • 5

Abstract

In this paper, we study the concentration properties of the kernel matrixassociated to a kernel function. Specifically, we derive new concentrationinequalities for the spectrum of an kernel matrix, quantifying its deviationwith respect to an associated integral operator. The main difference with mostresults in the literature is that we do not assume the positive definiteness ofthe kernel. Instead, we introduce Sobolev type hypothesis on the regularity ofthe kernel. We show how this is well suited to the study of kernels dependingonly on the distance between two points in a metric space, in which case theregularity depends only on the decay of the eigenvalues. This is connected withgeometric random graphs, which we study further, exhibiting explicit formulasfor the spectrum and its fluctuation.

 

Introduction (beta)

None

 

Conclusion (beta)

None