作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2012, Vol. 38 ›› Issue (21): 193-196. doi: 10.3969/j.issn.1000-3428.2012.21.052

• 人工智能及识别技术 • 上一篇    下一篇

基于PSO的BP神经网络学习算法

王爱平,江 丽   

  1. (安徽大学计算机科学与技术学院,合肥 230601)
  • 收稿日期:2012-01-04 出版日期:2012-11-05 发布日期:2012-11-02
  • 作者简介:王爱平(1956-),女,教授,主研方向:人工智能,数据库技术,容错控制;江 丽,硕士研究生
  • 基金资助:
    国家自然科学基金资助项目“熵理论在非高斯随机系统的故障诊断和容错控制中的应用”(61074071)、“非高斯随机分布控制系统的集成故障诊断与容错控制研究”(61104022)

BP Neural Network Learning Algorithm Based on Particle Swarm Optimization

WANG Ai-ping, JIANG Li   

  1. (School of Computer Science and Technology, Anhui University, Hefei 230601, China)
  • Received:2012-01-04 Online:2012-11-05 Published:2012-11-02

摘要: 针对标准反向传播(BP)算法收敛速度慢和易陷入局部极值等缺陷,提出一种基于粒子群优化的BP神经网络学习算法。采用标准BP梯度下降法调整权值,利用粒子群优化算法进行网络权值及阈值的修正。将该算法与标准BP算法及传统基于粒子群优化BP网络算法进行仿真比较。实验结果表明,该算法能够克服标准BP算法的缺点,性能优于其他2个BP网络优化模型。

关键词: 神经网络, 反向传播算法, 粒子群优化, 梯度下降法, 函数拟合

Abstract: For the standard Back Propagation(BP) algorithm usually has the limitations of slow convergence and local extreme values, a new BP neural network learning algorithm based on Particle Swarm Optimization(PSO) is proposed. The main idea of the model is to modify weight and threshold using PSO based on the weight adjustments of gradient descent method in BP algorithm. It evaluates the model by using simulation test of five typical complex functions and compares it with other two models including standard BP network and traditional PSO based BP network. Experimental results show that it can overcome the limitations of slow convergence and local extreme values for BP network and perform better than other two kinds of optimized BP network models.

Key words: neural network, Back Propagation(BP) algorithm, Particle Swarm Optimization(PSO), gradient descent method, function fitting

中图分类号: