Why Modified exponential covariance kernel is empirically successful

A theoretical explanation

authored by

Olga Kosheleva, Michael Beer

Abstract

It is known that in the first approximation, many real-life stationary stochastic processes are well- described by an exponential covariance kernel C(u) = exp(-a|u|). Empirical evidence shows that in many practical situations, a good second approximation is provided by the modified exponential covari- ance kernel C(u) = exp(-a |u|) (1-r|u|). In this paper, we provide a theoretical explanation for this empirical phenomenon.

External Organisation(s)

University of Texas at El Paso
University of Liverpool

Type

Article

Journal

Journal of Uncertain Systems

Volume

10

Pages

10-14

No. of pages

5

ISSN

1752-8909

Publication date

02.2016

Publication status

Published

Peer reviewed

Yes

ASJC Scopus subject areas

Computer Vision and Pattern Recognition, Control and Optimization, Artificial Intelligence

Electronic version(s)

http://www.worldacademicunion.com/journal/jus/jusVol10No1paper02.pdf (Access: Open)