期刊文献+

神经网络的两种结构优化算法研究 被引量:11

Two Structure Optimization Algorithms for Neural Networks
下载PDF
导出
摘要 提出了一种基于权值拟熵的“剪枝算法”与权值敏感度相结合的新方法,在“剪枝算法”中将权值拟熵作为惩罚项加入目标函数中,使多层前向神经网络在学习过程中自动约束权值分布,并以权值敏感度作为简化标准,避免了单纯依赖权值大小剪枝的随机性.同时,又针对剪枝算法在优化多输入多输出网络过程中计算量大、效率不高的问题,提出了一种在级联—相关(cascade-correlation,CC)算法的基础上从适当的网络结构开始对网络进行构建的快速“构造算法”.仿真结果表明这种快速构造算法在收敛速度、运行效率乃至泛化性能上都更胜一筹.* Based on pseudo-entropy of weights, a new method is proposed to integrate pruning algorithm with sensitivity weights. The pruning algorithm introduces the pseudo-entropy of weights as a penalty term into the normal objective function, and the distribution of weights is automatically constrained by a muhilayer feed-forward neural network during the training process. The weight sensitivity is served as the simplification criteria of pruning to avoid the pruning randomicity caused by only using the weights. Meanwhile, for the problems of heavy computation hurden and low efficiency of pruning algorithm in optimizing the multi-input and multi-output networks, a fast constructive algorithm is put forward, which is based on the Cascade-Correlation (CC) algorithm and constructs the new neural network from a proper network structure. The simulation results show that this fast constructive algorithm is a better choice in terms of convergence rate, computational efficiency and even generalization performance.
出处 《信息与控制》 CSCD 北大核心 2006年第6期700-704,710,共6页 Information and Control
基金 国家自然科学基金资助项目(60574051)
关键词 神经网络结构优化 剪枝算法 权值拟熵 权值敏感度 快速构造算法 泛化性能 neural network structure optimization pruning algorithm pseudo-entropy of weights weight sensitivity fast constructive algorithm generalization performance
  • 相关文献

参考文献10

  • 1鲍晓红,贾英民.用神经元网络辨识非线性系统中的网络结构选择[J].控制理论与应用,1997,14(4):489-494. 被引量:5
  • 2Erdogan S S,Ng G S,Chan P K H.Measurement criteria for neural network pruning[A].Proceedings of the 1996 IEEE Region 10 TENCON-Digital Signal Processing Applications Conference[C].Piscataway,NJ,USA:IEEE,1996.83~89.
  • 3Karninl E D.A simple procedure for pruning back-propagation trained neural networks[J].IEEE Transactions on Neural Networks,1990,1(2):239 ~242.
  • 4Babri H A,Kot A C,Tan N T,et al.Dynamic pruning algorithms for improving generalization of neural networks[A].Proceedings of the 1997 1st International Conference on Information,Communications and Signal Processing[C].Piscataway,NJ,USA:IEEE,1997.679~683.
  • 5Fahlman S E,Lebiere C.The cascade-correlation learning architecture[A].Advances in Neural Information Processing System 2[C].San Francisco,CA,USA:Morgan Kaufmann Publishers Inc.,1990.524~532.
  • 6Cover T M,Thomas J A.Elements of Information Theory[M].Peking:Tsinghua University Press,2003.12 ~ 15.
  • 7高大启.有教师的线性基本函数前向三层神经网络结构研究[J].计算机学报,1998,21(1):80-86. 被引量:241
  • 8李玉鉴.前馈神经网络中隐层神经元数目的一种直接估计方法[J].计算机学报,1999,22(11):1204-1208. 被引量:20
  • 9戴葵.神经网络设计[M].北京:机械工业出版社,2002.399-421.
  • 10Hamamoto M,Kamruzzaman J,Kumagai Y,et al.Generalization abilityof artificial neural network using Fahlman and Lebiere's learning algorithm[J].IEICE Transactions on Fundamentals of Electronics,Communications and Computer Sciences,1993,E76-A(2):242 ~247.

二级参考文献21

共引文献290

同被引文献88

引证文献11

二级引证文献66

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部