On the Regulatisation-enhanced Training of RBF Networks

Radial fhndions (RBF) network is an important common place in neum-fnzzy systems. In particolar, their common dversal approximator properties make hrzzy systems as well as n e o d network systems excellent representations for system modelling. In data-based modelling, it is important that, the overfitting should be avoided to eohana the generslisstion cspabaty of the model since tbh is BII oltimate performance m a n m for the validity of the model. In this mpect, in the majority of the reported rrserrrehes with RBF network, the issue of overfitting is omitted and modelling envm am endeavoured to vanish at the price of hidden degradation in genernlisation properties of the network. Tbe work addresses this issue in the RBF neoral networks for enbnnwd nenro-fuzzy system modelling.