Boosting the Performance of RBF Networks with Dynamic Decay Adjustment

Radial Basis Function RBF Networks also known as networks of locally tuned processing units see are well known for their ease of use Most algorithms used to train these types of net works however require a xed architecture in which the number of units in the hidden layer must be determined before training starts The RCE training algorithm introduced by Reilly Cooper and Elbaum see and its probabilistic extension the P RCE algorithm take advantage of a growing structure in which hidden units are only introduced when necessary The nature of these al gorithms allows training to reach stability much faster than is the case for gradient descent based methods Unfortunately P RCE networks do not adjust the standard deviation of their prototypes individually using only one global value for this parameter This paper introduces the Dynamic Decay Adjustment DDA al gorithm which utilizes the constructive nature of the P RCE al gorithm together with independent adaptation of each prototype s decay factor In addition this radial adjustment is class dependent and distinguishes between di erent neighbours It is shown that networks trained with the presented algorithm perform substan tially better than common RBF networks