Title :
Hopfield neural networks control for optimal solutions
Author :
Makki, Assaad ; Siy, Pepe
Author_Institution :
Dept. of Electr. & Comput. Eng., Wayne State Univ., Detroit, MI, USA
Abstract :
It is shown that introducing control elements into the Hopfield neural network aids in avoiding local minima in the network solution. Procedures for designing a modified Hopfield neural network with control elements are discussed. The application of the modified neural network to analog-to-digital (A/D) conversion is illustrated. It is shown how to determine the controllers to obtain the correct behavior of an A/D converter by utilizing the theory of variable structure systems. Also, it is shown that the solution presented by the output is independent of the initial state of the modified neural network
Keywords :
Hopfield neural nets; analogue-digital conversion; optimisation; variable structure systems; A/D converter; Hopfield neural networks; local minima; variable structure systems; Artificial neural networks; Biological neural networks; Computer networks; Control systems; Feedback control; Hopfield neural networks; Neurons; Optimal control; Parallel processing; Variable structure systems;
Conference_Titel :
Neural Networks, 1992. IJCNN., International Joint Conference on
Conference_Location :
Baltimore, MD
Print_ISBN :
0-7803-0559-0
DOI :
10.1109/IJCNN.1992.227301
