Optimising the Widths of Radial Basis Functions

In the context of regression analysis with penalised linear models (such as RBF networks) certain model selection criteria can be diierentiated to yield a re-estimation formula for the regularisation parameter such that an initial guess can be iteratively improved until a local minimum of the criterion is reached. In this paper we discuss some enhancements of this general approach including improved computational ee-ciency, detection of the global minimum and simultaneous optimisation of the basis function widths. The beneets of these improvements are demonstrated on a practical problem.