基于混沌变量的前向神经网络结构优化设计

Optimization design of feed-forward neural network structure based on chaos variables

提出一种关于多层前向神经网络结构的混沌优化设计方法。将混沌变量引入神经网络结构的优化搜索中,使得神经网络的隐层节点数以及所有权参数都处于混沌状态中,整个网络结构呈现为动态变化。从动态的神经网络结构中,根据性能指标来寻找一个全局最优或近似于全局最优的网络结构。仿真实验表明,采用该方案得到的神经网络结构模型对异或问题、非线性函数具有较高的逼近精度和较好的泛化能力。

The optimization design method is proposed for feed-forward neural network structure by means of chaos ergodicity and randomicity. Chaos variables are applied in searching for neural network structure, in which node numbers of hidden layer and all weight parameters of neural network are in chaotic state. All neural network structure is protean. A globally optimal or approximate globally optimal neural network structure is found according to performance standard from dynamic neural networks. Simulation results show feed-forward neural network has high approximation precision and good generalization capability to XOR problem and nonlinear function.