A unified approach to selecting optimal step lengths for adaptive vector quantizers

Author

Andrew, Lachlan L H ; Palaniswami, Marimuthu

Author_Institution

Dept. of Electr. & Electron. Eng., Melbourne Univ., Parkville, Vic., Australia

Volume

44

Issue

4

fYear

1996

fDate

4/1/1996 12:00:00 AM

Firstpage

434

Lastpage

439

Abstract

This paper presents expressions for the optimal step length to use when training a vector quantizer by stochastic approximation. By treating each update as an estimation problem, it provides a unified framework covering both batch and incremental training, which were previously treated separately, and extends existing results to the semibatch case. In addition, the new results presented provide a measurable improvement over results which were previously thought to be optimal

Keywords

adaptive signal processing; approximation theory; optimisation; stochastic processes; vector quantisation; adaptive vector quantizers; batch training; estimation problem; incremental training; optimal step lengths selection; semibatch training; stochastic approximation; update; vector quantizer training; Communications Society; Data compression; Distortion measurement; Mean square error methods; Nearest neighbor searches; Noise level; Probability density function; Quantization; Stochastic processes; Training data;

fLanguage

English

Journal_Title

Communications, IEEE Transactions on

Publisher

ieee

ISSN

0090-6778

Type

jour

DOI

10.1109/26.489089

Filename

489089