Camps Valls, GustavoMartínez Ramón, ManelRojo-Álvarez, José LuisMuñoz Marí, Jordi2009-02-042009-02-042007-01-011070-9908http://hdl.handle.net/10115/1910Nonlinear system identification based on relevance vector machines (RVMs) has been traditionally addressed by stacking the input and/or output regressors and then performing standard RVM regression. This letter introduces a full family of composite kernels in order to integrate the input and output information in the mapping function efficiently and hence generalize the standard approach. An improved trade-off between accuracy and sparsity is obtained in several benchmark problems. Also, the RVM yields confidence intervals for the predictions, and it is less sensitive to free parameter selection.enAtribución-NoComercial-SinDerivadas 3.0 Españahttp://creativecommons.org/licenses/by-nc-nd/3.0/es/Telecomunicacionesinfo:eu-repo/semantics/articleinfo:eu-repo/semantics/openAccess3325 Tecnología de las Telecomunicaciones