基于蚂蚁算法的RBF网络参数的两阶段优化

提出了一种径向基函数(RBF)网络中心参数仿生优化算法,该算法基于改进的蚂蚁算法分两阶段完成。第一阶段为:首先进行网络参数范围估算,然后根据一定的步长对网络参数区间取离散点,最后蚂蚁根据在各个离散点的信息素的多少来选择路径,从而进行网络参数优化。第二阶段为:利用第一阶段的结果进行局部区间扩张,从而进行进一步优化。用蚂蚁算法优化后的网络对典型的混沌时间序列进行预测,结果表明其预测的各项误差低于常规的优化方法。

Two Phase Optimization of RBF Neural Network Parameter Based on Ant Colony Algorithm

In the paper ant colony algorithm is improved and the parameters of radial basis function network are optimized in two phase.The first phase can be described as:First,the extent of every parameter is estimated;second,discrete points are got from every extent according to definite step;at last,ant colony selects the points according to the information quantity,so the network parameters are optimized;The second phase can be described as:The result of the first phase is locally expanded,so new extent is formed and can be optimized more precisely.The typical chaos sequence is predicted through the optimized network and the prediction result shows that the arrived error is far smaller than the corresponding part of the traditional algorithm.