Optimization of non-quadratic cost functions associated with Hopfield-like neural networks: impact of initial coefficient values

Author

Shoaib Bhuiyan, M. ; Iwahori, Yuji ; Iwata, A.

Author_Institution

Dept. of Electr. Eng. & Comput. Sci., Nagoya Inst. of Technol.

Volume

2

fYear

1999

fDate

1999

Firstpage

538

Abstract

An iterative algorithm to minimize the non-quadratic cost functionals associated with Hopfield-like neural networks is shown to reach a stable state faster (with fewer iterations required) and also to extract a better edge map when applied to edge extraction by periodically reducing the value of the coefficients of the non-quadratic cost functionals associated with Hopfield-like neural networks. The impact of the changes in the initial coefficient values and the impact of different patterns of the periodic step-down have been studied

Keywords

Hopfield neural nets; edge detection; minimisation; Hopfield-like neural networks; edge map extraction; initial coefficient values; iterative algorithm; nonquadratic cost functional minimisation; periodic step-down; stable state; Artificial neural networks; Computer networks; Computer vision; Cost function; Educational technology; Hopfield neural networks; Image edge detection; Image restoration; Neural networks; Stochastic processes;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Information Processing, 1999. Proceedings. ICONIP '99. 6th International Conference on

Conference_Location

Perth, WA

Print_ISBN

0-7803-5871-6

Type

conf

DOI

10.1109/ICONIP.1999.845651

Filename

845651