Realization of the Crossover-Type State Transition through the Intercommunication in a Group of Neural Networks, and Their Applications to Combinatorial Optimization Problems

Abstract

In genetic algorithm, which is one of the methods for solving combinatorial optimization problems, a better state is said to be able to emerge out of local optimal states by the crossover operation which means the genetic intercommunication among several individual states. This paper presents the fundamentals of 'crossover-type state transition in a group of neural networks. 'In practice, a system composed of several neural networks is assumed in which they communicate with one another and the automatical crossover-type state transition is realized by neurons' input and output relations. The authors verify the characteristics and effectiveness of the presented state transtion in comparison with genetic algorithm through simple combinatorial optimization problems,