Kernel method for nonlinear analysis: identification of a biological control system

Abstract Due to the high dimensionality of computation involved, the Volterra-Wiener kernel method has found only limited application in nonlinear system analysis. By means of a finite-dimensional approximation combined with the least-square error method, an efficient algorithm was developed for computing kernels of a nonlinear system subjected to random inputs. An application to a physiological control system demonstrates the practically of this algorithm as well as suggesting the prospective usage of the kernel method for nonlinear-system analysis in the future.