Convergence of Exponentiated Gradient Algorithms

Date

Authors

Hill, Simon

Williamson, Robert

Publisher

Institute of Electrical and Electronics Engineers (IEEE Inc)

Abstract

This paper studies three related algorithms: the (traditional) gradient descent (GD) algorithm, the exponentiated gradient algorithm with positive and negative weights (EG± algorithm), and the exponentiated gradient algorithm with unnormalized positive a

Keywords

Keywords: Algorithms; Convergence of numerical methods; Echo suppression; Mathematical models; Exponentiated gradient algorithms; Mean square error; Room acoustics; Acoustic signal processing Acoustic echo cancellation; Echo; Exponentiated gradient descent; Learning rate; Mean squared error; Room acoustics

URI

http://hdl.handle.net/1885/92748

Collections

ANU Research Publications

Source

IEEE Transactions on Signal Processing

Type

Journal article

Restricted until

2037-12-31

Downloads

File

Description

01_Hill_Convergence_of_Exponentiated_2001.pdf (151.91 KB)

02_Hill_Convergence_of_Exponentiated_2001.pdf (963.35 KB)