Optimal dimensioning of counter propagation neural networks


The absence of automated tools in the area of automated neural network design can be explained by the corresponding paucity of rigorous neural network composition techniques. The author suggests a hybrid architecture as the basis for a computer-aided neural network engineering tool. Such a tool is expected to complete automatically the minute yet important neural network architecture details. The author demonstrates the approach by developing an automatic counterpropagation neural network design module. It includes a mechanized Kohonen layer configurator, which combines A* and simulated annealing search techniques to achieve both automated dimensioning of the layer and simultaneous selection of its weights.

Published in: IJCNN-91-Seattle International Joint Conference on Neural Networks
Date of Conference: 08-12 July 1991
Date Added to IEEE Xplore06 August 2002
Print ISBN:0-7803-0164-1
DOI: 10.1109/IJCNN.1991.155376
Publisher: IEEE
Conference Location: Seattle, WA, USA