Skip to Main Content
A novel construction method of functional link artificial neural networks (FLANN) using support vector regression (SVR) was presented and applied to dynamic modeling for sensor. In this method, the SVR was compared with generic FLANN, which had a form similar to the SVR, and was shown to be equivalent to the FLANN estimate. An optimum FLANN solution could be obtained by SVR with the proper choice of the parameters. Therefore, the new FLANN could be uniquely obtained due to solving a quadratic programming instead of an iterative problem. The dynamic modeling experiment results show that, a generic FLANN has been developed to solve the same problem for comparison, the presented method is higher in accuracy and more robust in noise resistance.