Robust estimation of the structured covariance matrix for heavy-tailed distributions (2023)

sun, u, Babu, P. i Palomar, DP (2015).Robust estimation of the structured covariance matrix for heavy-tailed distributions. WIEEE 2015 International Conference on Acoustics, Speech and Signal Processing, ICASSP 2015 - Proceedings(σελ. 5693-5697). [7179062] (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings; Vol. 2015 - August). Institute of Electrical and Electronics Engineers Inc.https://doi.org/10.1109/ICASSP.2015.7179062

@in the process{5cac74378b3b4668a7a623201be7737e,

title = "Reliable estimation of a structured covariance matrix for heavy-tailed distributions",

abstract = "In this paper we consider a robust non-Gaussian covariance estimation problem. Specifically, the Tyler M estimator is used for samples drawn from a fat-tailed elliptic distribution. In some applications, the covariance matrix is ​​naturally structured. with the previous structure is beneficial for improving the estimation accuracy The problem is formulated as a minimization with the constraints of the Tyler cost function, where the construction is characterized by a set of constraints. is a convex programming sequence. set of matrices that can be decomposed into a sum of positive semidefinite matrices of first order, which has a wide range of applications, the algorithm is modified with much less complexity. Simulation results show that the proposed Tyler estimator with structural constraints achieves a smaller estimation error than the unconstrained case.",

autor = "Ying Sun i Prabhu Babu i Palomar, {Daniel P.}",

note = "Publisher Copyright: {\textcopyright} 2015 IEEE.; 40th IEEE International Conference on Acoustics, Speech and Signal Processing, ICASPP 2015; Conference Date: 4/19/2014 to 4/24/2014",

year = "2015",

month = August,

day = "4",

doi = "10.1109/ICASSP.2015.7179062",

language = "English (US)",

series = "ICASPP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",

εκδότης = "Institute of Electrical and Electronics Engineers Inc.",

pages = "5693--5697",

booktitle = "IEEE 2015 International Conference on Acoustics, Speech and Signal Processing, ICASP 2015 - Proceedings",

address = "United States",

}

Robust estimation of the structured covariance matrix for heavy-tailed distributions./Sun, YingBabu, Prabhu; Palomar, Daniel P.
IEEE 2015 International Conference on Acoustics, Speech and Signal Processing, ICASSP 2015 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2015. σελ. 10-10. 5693-5697 7179062 (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings; Volume 2015 - August).

Research achievements:Chapter in book/exhibition/conference proceedings›Contribution to the conference

TY-GEN

T1 - Robust estimation of the structured covariance matrix for heavy-tailed distributions

AU - Sunday, Ying

AU – Babu, Prabhu

UA – Palomar, Daniel P.

N1 - Publisher Copyright:© 2015 IEEE.

PY - 8/4/2015

Y1 - 2015/8/4

N2 – In this paper we consider the robust covariance estimation problem in a non-Gaussian system. Specifically, the Tyler's M estimator is used for samples drawn from an elliptical distribution with fat. In some applications, the covariance matrix has a naturally defined structure. Therefore, including prior structure information in the estimation process is beneficial to improve estimation accuracy. The problem is formulated as a constrained minimization of the Tyler cost function, where the construction is characterized by a set of constraints. A numerical algorithm based on major importance minimization is derived for general structures that can be characterized as a convex set, in which a convex programming sequence is solved. For the set of matrices that can be decomposed into the sum of first positive semidefinite matrices, which has a wide range of applications, the algorithm is modified with much less complexity. The simulation results show that the proposed Tyler estimator with structural constraints achieves a smaller estimation error than the unconstrained case.

AB - In this paper we consider the robust covariance estimation problem in a non-Gaussian system. Specifically, the Tyler's M estimator is used for samples drawn from an elliptical distribution with fat. In some applications, the covariance matrix has a naturally defined structure. Therefore, including prior structure information in the estimation process is beneficial to improve estimation accuracy. The problem is formulated as a constrained minimization of the Tyler cost function, where the construction is characterized by a set of constraints. A numerical algorithm based on major importance minimization is derived for general structures that can be characterized as a convex set, in which a convex programming sequence is solved. For the set of matrices that can be decomposed into the sum of first positive semidefinite matrices, which has a wide range of applications, the algorithm is modified with much less complexity. The simulation results show that the proposed Tyler estimator with structural constraints achieves a smaller estimation error than the unconstrained case.

UR — http://www.scopus.com/inward/record.url?scp=84946049874&partnerID=8YFLogxK

UR — http://www.scopus.com/inward/citedby.url?scp=84946049874&partnerID=8YFLogxK

U2 - 10.1109/ICASSP.2015.7179062

DO - 10.1109/ICASSP.2015.7179062

M3 - Contribution to the conference

AN-SCOPUS:84946049874

T3 - ICASPP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

SP-5693

EP-5697

BT - IEEE 2015 International Conference on Acoustics, Speech and Signal Processing, ICASP 2015 - Proceedings

PB - Institute of Electrical and Electronics Engineers S.A.

T2 - 40th International IEEE Conference on Acoustics, Speech and Signal Processing, ICASP 2015

Y2 - April 19, 2014 to April 24, 2014

IS -