自适应混合混沌神经网络及其在TSP中的应用

为了进一步提高混沌神经网络的收敛效果,将T.Kwok和K.A.Smith所总结的内、外两类混沌神经网络相结合,并加入自适应方法,提出了自适应混合混沌神经网络。这种网络可以同时通过线性和非线性两种途径来改变网络能量函数进行搜索,这使得搜索过程具有更为丰富的动力学行为。同时由于自适应方法的引入,保证能量函数的变化在整个搜索过程中对动态方程有较强的作用,从而减少了网络收敛时的迭代步数。仿真研究表明,在求解TSP上,其效果优于随机混沌模拟退火网络。

Self-adapt Hybrid Chaotic Neural Network and Its Application to TSP

In order to improve the convergent effect of chaotic neural network, the self-adapt hybrid chaotic neural network is introduced by combining the internal approach and external approach that are summarized by T. Kwok and K.A. Smith and adding self-adapt method. This network can change network energy in linear manner and nonlinear manner synchronously to searching results. This network makes searching of energy function have rich kinetic behaves. At the same time, because of adding self-adapt method, it is ensured that the changing of energy function has always strong effect on dynamic function. So the number of iteration is reduced. Simulations show that this network has better effect than SCSA, when using it to solve TSP.