This paper proposes a method for solving combinatorial optimization problems. The method uses Hopfield neural networks and a genetic algorithm to compensate each other's defects. The defect of the Hopfield neural network is capture by locally optimal solutions. The defect of genetic algorithms is the difficulty of coding. It is difficult to construct a coding scheme that does not generate unconstrained solutions in crossover and mutation operations.
In the proposed method, the Hopfield neural network and a genetic algorithm are used alternately. Solutions obtained with the Hopfield neural network are applied to the genetic algorithm to escape from locally optimal solutions. In the genetic algorithm, simple and direct coding is employed, and unconstrained solutions are repaired by the local search property of the Hopfield neural network. Simple and direct coding can be implemented easily and is effective for inheritance.
The method is evaluated by investigating three NP-complete problems: the maximum clique problem, the node cover problem, and the traveling salesman problem. © 1998 Scripta Technica, Syst Comp Jpn, 29(10): 68–74, 1998
[1]
Yoshiyasu Takefuji,et al.
A parallel improvement algorithm for the bipartite subgraph problem
,
1992,
IEEE Trans. Neural Networks.
[2]
John J. Grefenstette,et al.
Genetic Algorithms for the Traveling Salesman Problem
,
1985,
ICGA.
[3]
Behrooz Kamgar-Parsi,et al.
On problem solving with Hopfield neural networks
,
1990,
International 1989 Joint Conference on Neural Networks.
[4]
J J Hopfield,et al.
Neural networks and physical systems with emergent collective computational abilities.
,
1982,
Proceedings of the National Academy of Sciences of the United States of America.
[5]
Derek M. Wells,et al.
Solving degenerate optimization problems using networks of neural oscillators
,
1992,
Neural Networks.
[6]
Tal Grossman,et al.
Applying the INN model to the Maximum Clique problem
,
1993,
Cliques, Coloring, and Satisfiability.