摘要
为了提高径向基神经网络训练精度,提出一种混合优化算法。该算法利用粒子群优化算法全局搜索能力强的特点,避免了K均值算法受初始点选择的不利影响,提高了网络中心的搜索速度;同时采用动态权值算法避免径向基神经网络可能出现的病态问题,进一步提高网络的逼近能力。锅炉燃烧实例表明了改进算法的有效性和实用性。
In order to improve the training accuracy of radial basis neural network, thin paper proposed a hybrid optimization algorithm. The algorithm used the strong global search ability of Particle Swarm Optimization (PSO) algorithm to avoid the adverse effect by choosing initial point in the K-means algorithm, thus improving the network center search speed. Meanwhile, the dynamic weight algorithm was used to avoid the ill-posed problem, and to further improve the network approximation ability. The boiler combustion instance indicates that the improved algorithm is efficient and practical.
出处
《计算机应用》
CSCD
北大核心
2013年第6期1771-1773,1779,共4页
journal of Computer Applications
关键词
锅炉燃烧
粒子群优化算法
K均值算法
变梯度算法
boiler combustion
Particle Swarm Optimization (PSO) algorithm
K-means algorithm
conjugate gradientalgorithm