基于模拟退火遗传算法的RBF网络的优化

Parameter Optimization of RBF Neural Network Based on Genetic and Simulated Annealing Algorithm

提出了引入模拟退火的遗传算法对径向基函数(RBF)网络中心参数进行优化的算法,算法中选择实数编码,采用引入模拟退火过程的多点交叉和区域内随机波动的变异方法。用此算法作了两个仿真实验:一是对典型的混沌时间序列的预测,二是对被干扰了的图象进行去干扰。结果表明:这种基于模拟退火遗传算法对RBF网络参数的优化是行之有效的。

In the paper a new genetic optimization algorithm, which is based on simulated annealing algorithm, is put forward and parametera of RBF network are optimized. In this new genetic optimization algorithm real number coding and multi-location crossover are used ,and the mutation range is given by random .To prove the right of the new algo- rithm, two experiments are clone in MATLAB. One experiment is the forecast of chaos sequence and the other is the re- moving of noise data in digital image .The results of the two experiments proves： the new genetic algorithm is fight and the error of the network is less than the error in traditionally designed RBF network.