Learning non-linear time-scales with kernel γ-filters

NEUROCOMPUTING

卷:72

Article

Camps-Valls, Gustavo¹  Munoz-Mari, Jordi¹  Martinez-Ramon, Manel²  Requena-Carrion, Jesus³  Luis Rojo-Alvarez, Jose³ 

[1] Univ Valencia, Dept Elect Engn, Escola Tecn Super Engn, Valencia, Spain

[2] Univ Carlos III Madrid, Dept Teoria Senal & Comunicac, E-28903 Getafe, Spain

[3] Univ Rey Juan Carlos, Dept Teoria Senal & Comunicac, Madrid, Spain

关键词: Gamma filter;    Support vector machine;    Kernel;    Non-linear system identification;   

DOI  :  10.1016/j.neucom.2008.10.004

来源: Elsevier

【 摘 要 】

A family of kernel methods, based on the gamma-filter structure, is presented for non-linear system identification and time series prediction. The kernel trick allows us to develop the natural non-linear extension of the (linear) support vector machine (SVM) gamma-filter [G. Camps-Valls, M. Martinez-Ramon, J.L. Rojo-Alvarez, E. Soria-Olivas, Robust gamma-filter using support vector machines, Neurocomput. J. 62(12) (2004) 493-499.], but this approach yields a rigid system model without non-linear cross relation between time-scales. Several functional analysis properties allow us to develop a full, principled family of kernel gamma-filters. The improved performance in several application examples suggests that a more appropriate representation of signal states is achieved. (C) 2008 Elsevier B.V. All rights reserved.

【 授权许可】

Free