Comparison of the conjugate gradient method to the method of recursive least square in terms of convergence rate

Author

Choi, Seungwon ; Lee, Sang Seol

Author_Institution

Dept. of Telecommun. Eng., Hanyang Univ., Seoul, South Korea

fYear

1993

fDate

18-20 May 1993

Firstpage

905

Lastpage

908

Abstract

The convergence speed of the conjugate gradient method (CGM) is compared to that of the Kalman method. From simulations, the required computation time of CGM is about 79%-101% of that of the Kalman method. All the parameter values are set in both methods such that the performance of the receiving array is acceptable for land mobile communications. The main factor that causes the computation time to increase in the Kalman method is updating the inverse subcorrelation matrix. The main reason the CGM is superior to conventional algorithms is that the optimal weights are obtained in P steps in any signal environment

Keywords

Kalman filters; array signal processing; computational complexity; conjugate gradient methods; convergence of numerical methods; land mobile radio; least squares approximations; matrix inversion; recursive estimation; Kalman method; computation time; conjugate gradient method; convergence speed; inverse subcorrelation matrix; land mobile communications; optimal weights; performance; receiving array; recursive least square; simulations; Adaptive algorithm; Adaptive arrays; Convergence; Gradient methods; Least squares methods; Mobile communication; Sampling methods; Sensor arrays; User-generated content; Wideband;

fLanguage

English

Publisher

ieee

Conference_Titel

Vehicular Technology Conference, 1993., 43rd IEEE

Conference_Location

Secaucus, NJ

ISSN

1090-3038

Print_ISBN

0-7803-1267-8

Type

conf

DOI

10.1109/VETEC.1993.510957

Filename

510957