混合粒子群算法优化神经网络的研究  

The Research of Neural Network Based on Hybrid Particle Swarm Optimization

作　　者：田雨波[1] 潘朋朋[1] 

机构地区：[1]江苏科技大学电子信息学院,江苏镇江212003

出　　处：《微电子学与计算机》2011年第6期156-159,共4页Microelectronics & Computer

基　　金：船舶工业国防科技预研基金项目(10J3.5.2);江苏省高校自然科学基础研究项目(07KJB510032)

摘　　要：针对BP神经网络初始权阈值的确定所具有的随机性和各个隐含层神经元数的不确定性,通过利用混合粒子群优化算法来同时优化神经网络的初始权阈值和结构.首先通过混合粒子群优化算法来确定一个较好的搜索空间,然后在这个解空间里再通过BP算法对网络进行训练和学习,搜索出最优的网络结构和权阈值.通过Iris模式分类、Wine模式分类问题和广义异或问题来对该模型进行训练和测试,相比遗传算法等其他算法,该模型可以获得更高的正确识别率,结果表明此方法是可行的.Aiming at the random of the determination of initial weight and the number of neurons of the hidden layers for BP neural network,this paper utilizes hybrid particle swarm optimization to optimize the initial weight and structure for the neural network.It is used to get a better search space by hybrid particle swarm optimization firstly,and then the network is trained and learned in the solution space by BP.Thereby,it searches out the optimal network structure and weight.This model is used to train and test by the use of Iris pattern classification,Wine pattern classification and generalized XOR problem,compared with other algorithm such as genetic algorithm,it can get higher recognition rate.The result shows that the way of this paper is feasible.

关 键 词：混合粒子群 神经网络 分类 广义异或 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]