A comparison of two neural network architectures for vector quantization

Author

Naraghi-Pour, Mort ; Hedge, M. ; Bourge, Fabrice

Author_Institution

Dept. of Electr. & Comput. Eng., Louisiana State Univ., Baton Rouge, LA, USA

Volume

i

fYear

1991

fDate

8-14 Jul 1991

Firstpage

391

Abstract

The authors investigate the performance of two neural network architectures for vector quantization. The two architectures are the multilayer feedforward network and the Hopfield analog neural network. If is found that for the feedforward network to have reasonably good performance, the number of hidden units must be unrealistically high: exponential in the number of dimensions and codewords. For the Hopfield analog model, on the other hand, the number of processors required is equal to the number of codewords and the resulting performance is very close to the optimum mean squared error

Keywords

data compression; encoding; neural nets; Hopfield analog neural network; codewords; dimensions; hidden units; multilayer feedforward network; neural network architectures; optimum mean squared error; performance; vector quantization; Computer architecture; Costs; Distortion measurement; Encoding; Feedforward neural networks; Hopfield neural networks; Image storage; Multi-layer neural network; Neural networks; Vector quantization;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks, 1991., IJCNN-91-Seattle International Joint Conference on

Conference_Location

Seattle, WA

Print_ISBN

0-7803-0164-1

Type

conf

DOI

10.1109/IJCNN.1991.155209

Filename

155209