摘要
针对BP神经网络数值预测中结构不易确定、易陷入局部最小等问题,利用遗传算法,采用二进制编码方式优化网络结构,采用浮点数编码方式优化网络参数;利用粒子群算法进一步改善网络参数,减少算法耗时,提高预测精度。仿真验证了算法的有效性。
Based on the genetic algorithm and the particle swarm optimization algorithm, a method is presented to solve such problems in BP neural network as network structure indeterminacy and easy trapping in local minima. In the genetic algorithm, the binary encoding is put forward to optimize the network structure and the floating encoding is put forward to optimize the network parameters. By using the particle swarm optimization algorithm, BP network parameters are improved further. The simulation results show that the training time is reduced and the prediction accuracy is improved by using the method.
出处
《信息工程大学学报》
2016年第5期518-523,共6页
Journal of Information Engineering University
基金
国家科技重大专项资助课题(2014ZX03006003)
关键词
BP神经网络
数值预测
遗传算法
粒子群优化算法
结构优化
参数优化
BP neural network
numerical prediction
Genetic Algorithm
Particle Swarm Optimization
structure optimization
parameters optimization
