BP神经网络性能与隐藏层结构的相关性探究
PDF下载 (484)杨守建,陈 恳.BP神经网络性能与隐藏层结构的相关性探究[J].宁波大学学报(理工版),2013,26(01):48-52.DOI:
YANG Shou-jian,CHEN Ken.Effect of Multi-hidden-layer on Performance of BP Neural Network[J].Journal of Ningbo University(Natural Science & Engineering Edition),2013,26(01):48-52.DOI:
| Title: | Effect of Multi-hidden-layer on Performance of BP Neural Network |
| 作者: | 杨守建, 陈 恳 |
| Author(s): | YANG Shou-jian, CHEN Ken |
| 关键词: | BP神经网络; 隐藏层层数; 网络学习率; 多项式拟合 |
| Keywords: | BP neural network; the number of hidden layers; network learning rata; curve fitting |
| 分类号: | TP391.4 |
| 文献标识码: | A |
| 摘要: | 利用BP神经网络对多个给定的复杂非线性系统控制进行定量研究, 着重讨论了BP神经网络因隐藏层层数和网络学习率之间差异从而引起对复杂非线性系统控制性能上的影响. 通过实验数据的对比分析发现, BP神经网络隐藏层层数的递增与系统控制性能的提升并不成正相关性, 网络学习率的选取范围可控制在0~2.0之间, 具体参数因控制对象而异, 可采用分段调试和二分法运算以确定最佳网络学习率参数. |
| Abstract: | In this paper, quantitative study on a given complex nonlinear system is presented using BP neural network. The correlation is explored between the control performance on the given complex nonlinear system and the number of hidden layers and the difference of network learning rate. The analytical and test results well indicate that the improvement of control performance is not proportional to the increased number of hidden layers, and the optimal selection of neural learning rate is generally within 0-2. The specific value of learning rate depends on the specifically given system and the most suitable value range can be determined using sub-debugging method. |
| 参考文献 /References: | [1] 周开利, 康耀红. 神经网络模型及其MATLAB仿真[M]. 北京: 清华大学出版社, 2006:2. [2] 朱双东. 神经网络应用基础[M]. 沈阳: 东北大学出版社, 2000:3. [3] Wei C. Efficiency evaluation method of simulation system based on BP neural network[C]. Proceedings 2nd Interna- tional Conference on Networking and Distributed Com- puting, 2011:41-44. [4] Peng L, Bao L, Du C. The evaluation of scientific rea- soning ability based on BP neural network[J]. Commu- nications in Computer and Information Science, 2011, 163:133-139. [5] Izeboudjen N, Bouridane A, Farah A, et al. Application of design reuse to artificial neural networks: case study of the back propagation algorithm[J]. Neural Computing and Applications, 2011, 21:1-14. [6] 张国翊, 胡铮. 改进BP 神经网络模型及其稳定性分析[J]. 中南大学学报:自然科学版, 2011, 42(1):115-124. [7] 毛永毅, 周康磊. 基于BP神经网络的定位跟踪算法[J].计算机工程与应用, 2011, 47(20):238-240. [8] 车建国, 耿学和. 基于改进算法的BP神经网络在故障诊断中的应用[J]. 战术导弹控制技术, 2011, 28(1):45- 47. [9] Li Y, Tian Y, Chen W. Multi-pattern recognition of sEMG based on improved BP neural network algorithm[C]. Proceedings of the 29th Chinese Control Conference, 2010:2867-2872. [10] Meng X, Meng X. Nonlinear system simulation based on the BP neural network[C]. Proceedings 3rd International Conference on Intelligent Networks and Intelligent Sys- tems, 2010:334-337. [11] Dong C, Dong L, Yang M. The application of the BP neural network in the nonlinear optimization[J] Advances in Intelligent and Soft Computing, 2011, 78:727-732. [12] 张华节, 李璀. 学习速率连续变化的BP神经网络算法[J]. 四川兵工学报, 2011, 32(4):61-64. |
| 备注/Memo: | 收稿日期: 2012?09?08.基金项目: 宁波市自然科学基金（2010A610109）.第一作者: 杨守建（1989－）, 男, 浙江长兴人, 在读硕士研究生, 主要研究方向: 电气自动化和人工智能. E-mail: 1233537@tongji.edu.cn.*通信作者: 陈 恳（1962－）, 男, 重庆人, 博士/副教授, 主要研究方向: 图像与视频处理、人工智能. E-mail: chenken@nbu.edu.cn 宁波大学学报（理工版）网址：http://journallg.nbu.edu.cn/ |
