Abstract

This paper addresses a kernel-based learning problem for a network of agentslocally observing a latent multidimensional, nonlinear phenomenon in a noisyenvironment. We propose a learning algorithm that requires only mild a prioriknowledge about the phenomenon under investigation and delivers a model withcorresponding non-asymptotic high probability error bounds. Both non-asymptoticanalysis of the method and numerical simulation results are presented anddiscussed in the paper.

 

Quick Read (beta)