期刊文献+

BP网络过拟合满足的不确定关系式 被引量:1

Uncertainty Relation Satisfied With Overfitting of BP Neural Network
下载PDF
导出
摘要 为了研究BP网络的过拟合现象 ,通过类比信息传递过程中的一般测不准关系式 ,建立了BP网络出现过拟合现象时的网络学习能力与推广能力之间满足的一般不确定关系式。通过模拟多种不同类型复杂程度函数的数值试验 ,确定出不确定关系式中的过拟合参数的取值范围为 1× 10 -5~ 5× 10 -4。根据一般不确定关系式 ,给出应用BP网络对给定样本集的训练过程中 ,避免出现过拟合现象 ,提高网络推广能力的方法。 In order to study the overfitting of BP neural network, a general uncertainty relation between the change of weighted value representing learning ability and the discrimination error of untraining sample sets representing generalization ability is revealed in the modeling of BP neural network by the analogy of general uncertainty relation in information transmitting process. Tests of numerical simulation for various of complicated functions are carried out to determine the value distribution (1×10\{-5\}~5×10\{-4\})of overfitting parameter in the uncertainty relation. Based on the uncertainty relation, the overfitting in the training process of given sample sets using BP network can be avoided.\;
作者 李祚泳 蔡辉
出处 《系统工程与电子技术》 EI CSCD 北大核心 2002年第9期94-96,125,共4页 Systems Engineering and Electronics
基金 国家"九五"重点科技攻关项目资助课题 (96-911-0 8-0 3 -0 5 )
关键词 BP网络 过拟合 不确定关系式 学习能力 推广能力 信息论 神经网络 BP network Overfitting Learning ability Generalization ability Uncertainty relation Information theory
  • 相关文献

参考文献8

  • 1Atiya A, Ji C Y. How Initial Conditions Affect Generalization Performance in Large Networks[J]. IEEE Trans. on Neural Networks, 1997, 8(2): 448-451.
  • 2Levin E, Tishiby N, Solia S A. A Statistical Approach to Learning and Generalization in Layered Neural Networks[J]. Proc. of the IEEE, 1990, 78(10): 1568-1574.
  • 3Murata N, Yoshizawa S, Amari S. Network Information Criterion-Determining the Number of Hidden Units for an Artificial Neural Network Model[J]. IEEE Trans. on Neural Networks, 1994, 5(6): 865-872.
  • 4Schwartz D B. Exhaustive Learning[J]. Neural Computation, 1990, 2: 374-385.
  • 5Wolpert D H. A Mathematical Theory of Generalization (Part I, Part II)[J]. Complex Systems, 1990, 4: 151-200; 201-249.
  • 6彭汉川 甘强 韦钰.提高前馈神经网络推广能力的若干实际方法[J].计算机工程与应用,1999,(1):47-48.
  • 7阎平凡.对多层前向神经网络研究的进一步看法[J].电子学报,1999,27(5):82-85. 被引量:25
  • 8查有梁.INFORMATION UNCERTAINTY PRINCIPLE[J].Chinese Science Bulletin,1989,34(1):86-87. 被引量:1

二级参考文献10

共引文献27

同被引文献13

  • 1张鸿宾.训练多层网络的样本数问题[J].自动化学报,1993,19(1):71-77. 被引量:23
  • 2阎平凡.人工神经网络的容量、学习与计算复杂性[J].电子学报,1995,23(5):63-67. 被引量:82
  • 3Baum E B,Haassler D.What size net gives valid generalization[A].NIPSI [C] .San Mateo,CA, 1989.81-90.
  • 4Moody J E.The effective nuber of parameters:An analysis of generalization and regulrization in nonlinear learning system [ A ]. NIPS 4[ C]. San Mateo, CA, 1992.847 -854.
  • 5Barron A R. Approximation and estimation bounds for artificial neural networks [J] .Machine Learning, 1994, (14) : 115 - 133.
  • 6Geman S. Neural networks and bias/variance dilemma [ J ]. Neural Computation, 1992, (4) : 1 - 58.
  • 7Cataltepe Z, Abu-mostafa Y. S, Magdon-Ismail M. No free lunch for earlystopping [J] .Neural Computation, 1999, (11):995- 1009.
  • 8Amari S, Murata N, Muller K R, et al. A symptotic statistical theory of overtraining and cross-validation [ J ]. IEEE Trana Neural Networks,1997,8(5) :985 - 996.
  • 9彭汉川 甘强 韦钰.提高前馈神经网络推广能力的若干实际方法[J].计算机工程与应用,1999,(1):47-48.
  • 10江学军,唐焕文.前馈神经网络泛化性能力的系统分析[J].系统工程理论与实践,2000,20(8):36-40. 被引量:44

引证文献1

二级引证文献9

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