Sensitivity based Growing and Pruning method for RBF network in online learning environments

How to define the architecture of classifiers dynamically is one of the major research topics in online learning. This paper presents a new online learning algorithm for Radial Basis Function Network named Sensitivity Based Neurons Growing and Pruning Method for RBF network (SBGAP). The performance of SBGAP is evaluated experimentally by comparing accuracy and the number of neurons with the existing methods. The experimental results show that SBGAP achieve litter higher accuracy with fewer hidden units in most situations.