Decorrelation of Neutral Vector Variables: Theory and Applications

Zhanyu Ma, Jing-Hao Xue, Arne Leijon, Zheng-Hua Tan, Zhen Yang, Jun Guo

Research output: Contribution to journalJournal articleResearchpeer-review

69 Citations (Scopus)

Abstract

In this paper, we propose novel strategies for neutral vector variable decorrelation. Two fundamental invertible transformations, namely, serial nonlinear transformation and parallel nonlinear transformation, are proposed to carry out the decorrelation. For a neutral vector variable, which is not multivariate-Gaussian distributed, the conventional principal component analysis cannot yield mutually independent scalar variables. With the two proposed transformations, a highly negatively correlated neutral vector can be transformed to a set of mutually independent scalar variables with the same degrees of freedom. We also evaluate the decorrelation performances for the vectors generated from a single Dirichlet distribution and a mixture of Dirichlet distributions. The mutual independence is verified with the distance correlation measurement. The advantages of the proposed decorrelation strategies are intensively studied and demonstrated with synthesized data and practical application evaluations.

Original languageEnglish
JournalI E E E Transactions on Neural Networks and Learning Systems
Volume29
Issue number1
Pages (from-to)129-143
Number of pages14
ISSN2162-237X
DOIs
Publication statusPublished - Jan 2018

Fingerprint

Principal component analysis

Cite this

Ma, Zhanyu ; Xue, Jing-Hao ; Leijon, Arne ; Tan, Zheng-Hua ; Yang, Zhen ; Guo, Jun. / Decorrelation of Neutral Vector Variables : Theory and Applications. In: I E E E Transactions on Neural Networks and Learning Systems. 2018 ; Vol. 29, No. 1. pp. 129-143.
@article{a9de112ff9b440418b56cd6913c3ee54,
title = "Decorrelation of Neutral Vector Variables: Theory and Applications",
abstract = "In this paper, we propose novel strategies for neutral vector variable decorrelation. Two fundamental invertible transformations, namely, serial nonlinear transformation and parallel nonlinear transformation, are proposed to carry out the decorrelation. For a neutral vector variable, which is not multivariate-Gaussian distributed, the conventional principal component analysis cannot yield mutually independent scalar variables. With the two proposed transformations, a highly negatively correlated neutral vector can be transformed to a set of mutually independent scalar variables with the same degrees of freedom. We also evaluate the decorrelation performances for the vectors generated from a single Dirichlet distribution and a mixture of Dirichlet distributions. The mutual independence is verified with the distance correlation measurement. The advantages of the proposed decorrelation strategies are intensively studied and demonstrated with synthesized data and practical application evaluations.",
author = "Zhanyu Ma and Jing-Hao Xue and Arne Leijon and Zheng-Hua Tan and Zhen Yang and Jun Guo",
year = "2018",
month = "1",
doi = "10.1109/TNNLS.2016.2616445",
language = "English",
volume = "29",
pages = "129--143",
journal = "I E E E Transactions on Neural Networks and Learning Systems",
issn = "2162-237X",
publisher = "IEEE",
number = "1",

}

Decorrelation of Neutral Vector Variables : Theory and Applications. / Ma, Zhanyu; Xue, Jing-Hao; Leijon, Arne; Tan, Zheng-Hua; Yang, Zhen; Guo, Jun.

In: I E E E Transactions on Neural Networks and Learning Systems, Vol. 29, No. 1, 01.2018, p. 129-143.

Research output: Contribution to journalJournal articleResearchpeer-review

TY - JOUR

T1 - Decorrelation of Neutral Vector Variables

T2 - Theory and Applications

AU - Ma, Zhanyu

AU - Xue, Jing-Hao

AU - Leijon, Arne

AU - Tan, Zheng-Hua

AU - Yang, Zhen

AU - Guo, Jun

PY - 2018/1

Y1 - 2018/1

N2 - In this paper, we propose novel strategies for neutral vector variable decorrelation. Two fundamental invertible transformations, namely, serial nonlinear transformation and parallel nonlinear transformation, are proposed to carry out the decorrelation. For a neutral vector variable, which is not multivariate-Gaussian distributed, the conventional principal component analysis cannot yield mutually independent scalar variables. With the two proposed transformations, a highly negatively correlated neutral vector can be transformed to a set of mutually independent scalar variables with the same degrees of freedom. We also evaluate the decorrelation performances for the vectors generated from a single Dirichlet distribution and a mixture of Dirichlet distributions. The mutual independence is verified with the distance correlation measurement. The advantages of the proposed decorrelation strategies are intensively studied and demonstrated with synthesized data and practical application evaluations.

AB - In this paper, we propose novel strategies for neutral vector variable decorrelation. Two fundamental invertible transformations, namely, serial nonlinear transformation and parallel nonlinear transformation, are proposed to carry out the decorrelation. For a neutral vector variable, which is not multivariate-Gaussian distributed, the conventional principal component analysis cannot yield mutually independent scalar variables. With the two proposed transformations, a highly negatively correlated neutral vector can be transformed to a set of mutually independent scalar variables with the same degrees of freedom. We also evaluate the decorrelation performances for the vectors generated from a single Dirichlet distribution and a mixture of Dirichlet distributions. The mutual independence is verified with the distance correlation measurement. The advantages of the proposed decorrelation strategies are intensively studied and demonstrated with synthesized data and practical application evaluations.

U2 - 10.1109/TNNLS.2016.2616445

DO - 10.1109/TNNLS.2016.2616445

M3 - Journal article

VL - 29

SP - 129

EP - 143

JO - I E E E Transactions on Neural Networks and Learning Systems

JF - I E E E Transactions on Neural Networks and Learning Systems

SN - 2162-237X

IS - 1

ER -