DocumentCode :
2901087
Title :
Optoelectronic radial basis function network training techniques
Author :
Waddie, Andrew J. ; Taghizadeh, Mohammad R.
Author_Institution :
Sch. of Eng. & Phys. Sci., Heriot-Watt Univ., Edinburgh, UK
fYear :
2005
fDate :
12-17 June 2005
Firstpage :
478
Abstract :
The radial basis function network (RBFN) is a multi-layer feed-forward neural network consisting of simple processing elements (neurons) with weighted interconnections between the elements. The weighted interconnections between the neurons can be calculated by any one of a wide range of optimisation algorithms. We will demonstrate that although the gradient descent method produces a "better" solution to the sample problem, limiting the number of allowable interconnection weights enhances the tolerance of the network to errors in the DOE reconstruction. In addition, the genetic algorithm optimisation method is significantly faster than the gradient descent method for the same number of allowable weight values.
Keywords :
diffractive optical elements; genetic algorithms; optical interconnections; optical neural nets; optoelectronic devices; radial basis function networks; DOE reconstruction; genetic algorithm; multi-layer feed-forward neural network; optimisation; radial basis function network; weighted interconnections; Digital signal processing; Feedforward neural networks; Feedforward systems; Genetic algorithms; Multi-layer neural network; Neural networks; Neurons; Radial basis function networks; US Department of Energy; Vertical cavity surface emitting lasers;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Lasers and Electro-Optics Europe, 2005. CLEO/Europe. 2005 Conference on
Print_ISBN :
0-7803-8974-3
Type :
conf
DOI :
10.1109/CLEOE.2005.1568256
Filename :
1568256
Link To Document :
بازگشت