摘要
提出一种基于动态粒子群算法的神经网络训练方法。神经网络权值选择是否合适直接关系到其非线性拟合能力,通过引入动态粒子群算法对神经网络进行训练,对神经网络各层连接权值进行优化。经过函数测试表明,相比粒子群算法,动态粒子群算法收敛速度更快且不易陷入局部,能更快更合理地训练神经网络从而优化网络连接权值。
A Neural Network algorithm based on dynamic weight has been proposed in this paper. The choose of Neural Network is directly determining the nonlinear fitting in controlling systems.The introduction of the Dynamic Particle Swarm Optimization in this article has optimized the connecting weight of Neural Network. The research indicates that the Dynamic Particle Swarm Optimization has the faster rate of convergence and the better effect for training Neural Network. The algorithm has the ability to avoid getting the minimum of the weight.
出处
《计算机安全》
2014年第4期5-8,共4页
Network & Computer Security
关键词
神经网络
粒子群算法
连接权值
动态权重
Neural Network
Particle Swarm Optimization
connecting weight
dynamic weight
