Hopfield-type neural networks with fuzzy sets to gather the convergent speed

To solve combinatorial optimization problems with Hopfield-type neural networks, the slope of the sigmoid function must be adjusted to a desirable narrow range. It was reported that the desirable range could be widened by changing the parameters of the energy function and the sigmoid function dynamically. Fuzzy parameters are introduced to perform scheduling of the networks. A fuzzy rule is proposed for the purpose of fast convergence while keeping the ability to minimize energy. Simulation results show its validity on the traveling salesman problem.<
>