Improving the generalisation properties of radial basis function neural networksFunded under: FP2-FUSION 10C
An important feature of radial basis function neural networks is the existence of a fast, linear learning algorithm in a network capable of representing complex non-linear mappings. Satisfactory generalisation in these networks requires that the network mapping be sufficiently smooth. It is shown that a modification to the error function allows smoothing to be introduced without significantly affecting the speed of training. A simple example is used to demonstrate the resulting improvement in the generalisation properties of the network.
Bibliographic Reference: Report: AEA FUS 94 EN (1991) 11 pp.
Availability: Available from the Librarian, UKAEA, Harwell Laboratory, Didcot, Oxon. OX11 ORA (GB)
Record Number: 199110609 / Last updated on: 1994-12-02
Original language: en
Available languages: en