一种混沌神经网络及其在优化计算中的应用

A Kind of Chaotic Neural Network and Its Application for Nonlinear Optimization

研究了一种具有混沌特性的神经网络 ,该网络具有瞬态混沌响应 ,类似于Hopfield网络的结构 ,但有比Hopfield网络更加丰富的动力学特征、更强的全局搜索能力。通过把混沌动力学与收敛动力学相结合 ,使网络逐渐由混沌神经网络向Hopfield网络过渡 ,达到控制混沌的目的 ,并且提供一个在全局最优解附近的初值 ,有效地解决了Hopfield网络的局部极值问题。该网络模型可以用来解决复杂的非线性优化问题。

A neural network model with chaotic character is studied. The network's structure is similar with that of Hopfield neural network, but it has richer and more flexible dynamics than Hopfield neural network, namely has only point attractors, so that it has higher ability of searching global optimal or near-optimal solutions. By combining chaotic dynamics and converging dynamics together, we make the neural network transit gradually to Hopfield network. By introducing converging factor, we get the aim of controlling chaos, provide initial value of Hopfield network that is near to the global optimal solutions, and solve the problem of local minimum. A great quantity of experiences show that the model can solve the complex nonlinear optimal problem.