一种改进混沌神经网络及其在组合优化问题中的应用

在传统的混沌神经网络模型基础上,提出了一种改进的混沌神经网络(ICNN) .通过引入时变的输出函数增益和修正自反馈系数的表达式,使该模型可有效地控制Sigmoid输出函数图形的陡度和模型演化中混沌动态的收敛过程,从而拥有更丰富的神经动力学特性与初值鲁棒性.该模型可有效地解决一系列组合优化问题(COP) ,解决了10个与4 8个城市的TSP问题

A kind of improved chaotic neural network and its application in combinatorial optimization problems

Based on the principle of traditional chaotic neural network model, a new improved chaotic neural network (ICNN) is presented. By introducing the decaying steepness parameter of the output function into this model and modifying the function of self-feedback connection weight, the proposed model can effectively take control of the slope of Sigmoid threshold function and the chaotic dynamic convergence of time evolutions, which leads to richer and more flexible dynamics and robustness of initial conditions. This model can availably solve kinds of combinatorial optimization problems (COP), the 10-city and 48-city TSP are figured out successfully in this paper.