[1]张米娜,韩红桂,乔俊飞.前馈神经网络结构动态增长-修剪方法[J].智能系统学报,2011,6(2):101-106.
ZHANG Mina,HAN Honggui,QIAO Junfei.Research on dynamic feedforward neural network structure based on growing and pruning methods[J].CAAI Transactions on Intelligent Systems,2011,6(2):101-106.
点击复制
ZHANG Mina,HAN Honggui,QIAO Junfei.Research on dynamic feedforward neural network structure based on growing and pruning methods[J].CAAI Transactions on Intelligent Systems,2011,6(2):101-106.
前馈神经网络结构动态增长-修剪方法
《智能系统学报》[ISSN 1673-4785/CN 23-1538/TP] 卷:
6
期数:
2011年第2期
页码:
101-106
栏目:
学术论文—机器学习
出版日期:
2011-04-25
- Title:
- Research on dynamic feedforward neural network structure based on growing and pruning methods
- 文章编号:
- 1673-4785(2011)02-0101-06
- Keywords:
- adaptive growing and pruning (AGP); BOD softmeasurement; neural network; self organization
- 分类号:
- TP183
- 文献标志码:
- A
- 摘要:
- 针对前馈神经网络隐含层神经元不能在线调整的问题，提出了一种自适应增长修剪算法(AGP)，利用增长和修剪相结合对神经网络隐含层神经元进行调整，实现神经网络结构的自组织，从而提高神经网络的性能.同时，将该算法应用于污水处理生化需氧量(BOD)软测量，仿真实验结果表明，与其他自组织神经网络相比，AGP具有较好的泛化能力及较高的拟合精度，能够实现出水BOD的预测.
- Abstract:
- Due to the unchangable online problem of hidden neurons in feedforward neural networks, an adaptive growing and pruning algorithm (AGP) was presented in this paper. This algorithm can insert and prune hidden neurons during the training process to adjust the structure of the network and achieve self organization of neural network structure, which can improve the performance of the neural network. Additionally, this algorithm has been applied to the biochemical oxygen demand (BOD) soft measurement of the wastewater treatment process. Experimental results show that the proposed algorithm can forecast the effluent BOD with better generalization ability and higher accuracy than other selforganizing neural networks.
- 参考文献/References:
-
［1］乔俊飞, 张颖. 一种多层前馈神经网络的快速修剪算法［J］. 智能系统学报, 2008, 3(2): 173176.
QIAO Junfei, ZHANG Ying. Fast unit pruning algorithm for multilayer feedforward network design［J］.CAAI Transactions on Intelligent Systems, 2008, 3(2): 173176.
［2］杨慧中,王伟娜. 神经网络的两种结构优化算法研究［J］. 信息与控制, 2006, 35(6): 700704.
YANG Huizhong, WANG Weina. Two structure optimization algorithms for neural networks［J］. Information and Control, 2006, 35(6): 700704.
［3］BORTRMAN M, ALADIEM M. A growing and pruning method for radial basis function networks［J］. IEEE Transaction on Neural Networks, 2009, 20(6): 10391045.
［4］HASSIBI B， STORK D G. Second order derivatives for network pruning: optimal brain surgeon［C］//Advances in Neural Information Processing Systems. San Mateo, USA: Morgan Kauffman, 1993： 164171.
［5］LAURET P, FOCK E, MARA T A. A node pruning algorithm based on a Fourier amplitude sensitivity test method［J］. IEEE Transactions on Neural Networks, 2006, 17 (2): 273293.
［6］ XU Jinhua, DANIEL W. A new training and pruning algorithm based on node dependence and Jacobian rank deficiency［J］. Neurocomputing, 2006, 70(1): 544558.
［7］MARSLAND S, SHAPIRO J. A selforganizing network that grows when required［J］. Neural Networks, 2002, 15(8):10411058.
［8］ISLAM M, SATTAR A, AMIN F, YAO Xin. A new adaptive merging and growing algorithm for designing artificial neural networks［J］. IEEE Trans Systems, Man, CyberneticsPart B: Cybernetics, 2009, 39(3): 705722.
［9］QIAO Junfei, HAN Honggui. A repair algorithm for RBF neural network and its application to chemical oxygen demand modeling［J］. International Journal of Neural Systems, 2010, 20(1): 6374.
［10］JOHNSON C, VENAYAGAMOORTHY G K, MITRA P. Comparison of a spiking neural network and an MLP for robust identification of generator dynamics in a multimachine power system［J］. Neural Networks, 2009, 22(5/6): 833841.
［11］ RUMELHART D E, HINTON G E,WILLIAMS R J. Learning internal representations by error propagation［M］.Parallel Distributed Processing. Cambridge,USA: MIT Press, 1986： 318362.
［12］韩红桂, 甑博然, 乔俊飞. 动态结构优化神经网络及其在溶解氧控制中的应用［J］. 信息与控制, 2010, 39(3): 354360.
HAN Honggui, ZHEN Boran, QIAO Junfei.Dynamic structure optimization neural network and its applications to dissolved oxygenic (DO) control［J］. Information and Control, 2010, 39(3): 345360.
［13］PENG Jianxun,LI Kang, HUANG Deshuang. A hybrid forward algorithm for RBF neural network construction［J］. IEEE Transactions on Neural Networks, 2006, 17 (6): 14391451.
备注/Memo
收稿日期：2010-04-22.
基金项目：国家“863” 计划资助项目(2007AA04Z160)；国家自然科学基金资助项目(60873043);北京市自然科学基金资助项目(4092010); 高等学校博士点专项科研基金资助项目(200800050004) .
通信作者：张米娜.
E-mail:zhang.mi.na@163.com.
作者简介：
张米娜，女，1986年生，硕士研究生，主要研究方向为神经网络结构优化设计、智能控制理论与应用.
韩红桂，男，1983年生，博士研究生，主要研究方向为智能信息处理、智能控制理论与应用.
乔俊飞，男，1968年生，教授，博士，主要研究方向为神经网络结构分析与设计、计算智能与智能优化控制.
更新日期/Last Update:
2011-05-19