A controller based on the energy function to improve convergence of a neural network
The feedback (Hopfield) model of a neural network is enhanced by the addition of a controller. This controller selectively chooses responses in the feedback loop to improve the performance of the recall operation. The process of selection is based on finding the response with the lowest (deepest) en...
Saved in:
Published in | IEEE Proceedings of the SOUTHEASTCON '91 pp. 741 - 745 vol.2 |
---|---|
Main Author | |
Format | Conference Proceeding |
Language | English |
Published |
IEEE
1991
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | The feedback (Hopfield) model of a neural network is enhanced by the addition of a controller. This controller selectively chooses responses in the feedback loop to improve the performance of the recall operation. The process of selection is based on finding the response with the lowest (deepest) energy at each step of iteration. Results of computer simulations are presented. The recall improvements for probes with any Hamming distance from the stored memories as well as for those with only one Hamming distance away from the memories are investigated. The effects of this controller on the performance of the model as the number of stored memories increases are studied.< > |
---|---|
ISBN: | 9780780300330 0780300335 |
DOI: | 10.1109/SECON.1991.147856 |