期刊文献+

尺度可调的核函数模型在震荡信号表示中的应用

Non-flat Function Representation with Scale-tunable Kernel Model
下载PDF
导出
摘要 传统固定尺度的核函数模型不适合稀疏地表示震荡信号.为了提高震荡信号表示的稀疏性,提出了一种尺度可调的核函数模型的建立方法.该方法通过正交最小二乘算法进行逐步回归建模,选择每一个回归子时,利用群搜索算法优化残差目标函数,计算相应的核函数的尺度.实验结果表明,可调核函数模型比传统的固定尺度核函数模型具有更强的稀疏性和泛化能力. Based on orthogonal least square algorithm(OLS),a new type of scale-tunable kernel model is proposed to enhance the sparseness of representation for non-flat functions.At each regressor stage,the scale of each term is selected by minimizing the residual using group search optimizer.Experimental results indicate that the new method can produce a kernel function model with better sparseness and generalization than the traditional fix-scale kernel models.
出处 《小型微型计算机系统》 CSCD 北大核心 2011年第10期2114-2117,共4页 Journal of Chinese Computer Systems
基金 国家自然科学基金项目(60672049 60773167 90920005 11026145)资助 国家″十一五″科技支撑计划课题项目(2006BAK11B03)资助 中央高校科研业务费资助 湖北省自然科学基金项目(2010CDB04205)资助
关键词 群搜索优化子 核模型 正交最小二乘 group search optimizer(GSO) kernel model orthogonal least squares(OLS)
  • 相关文献

参考文献8

  • 1Smola A. Regression estimation with support vector learning ma- chines[ D]. Technische University M{inchen, 1996. 被引量:1
  • 2Smola A, Scholkopf B, Ratsch G. Linear programs for automatic accuracy control in regression[ C]. Proceedings of the Ninth International Conference on Artificial Neural Networks, London, 1999, 575-580. 被引量:1
  • 3Chen S ,Billings S A,Luo W. Orthogonal least squares methods and their application to non-lincar system identification [ J ]. Int. J Control, 1989, 50(5) : 1873-1896. 被引量:1
  • 4He S, Wu Q H, Saunders J R. Group search optimizer: an optimi- zation algorithm inspired by animal searching behavior[ J]. IEEE Trans. on Evolutionary Computation, 2009, 13 (5) :973-990. 被引量:1
  • 5Vicente J, Lionello P, Emili B. Two-variable linear regression: modeling with orthogonal least-squares qnalysis [J]. J. Chem. Educ. , 2010, 87 (9) : 994-995:. 被引量:1
  • 6Mendez G, Hemandez M. Hybrid learning for interval type-2 fuzzy logic systems based on orthogonal least-squares and back-propaga- tion methods [J]. Information Sciences, 2009, 179 (13) : 2146- 2157. 被引量:1
  • 7Rajah L J, Thomas R F. Comparison on of generalized ganssian and laplacian modeling in DCT image coding [Jr]. IEEE Trans. on Signal Processing Letters, 1995, 2(5) : 81-82. 被引量:1
  • 8Silverman B W. Some aspects of the spine smoothing approach to non-parametric regression curve fitting[J].Journal of Royal Statistical Society, 1985,47 ( 1 ) : 1-52. 被引量:1

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