Improved pruning strategy for radial basis function networks with dynamic decay adjustment

Dynamic decay adjustment (DDA) is a fast algorithm to construct radial basis function (RBF) networks for classification problems. It is known that, despite its interesting features, DDA produces classifiers with high complexity, especially for large datasets. In this Letter a simple method to overcome this problem is proposed, which eliminates redundant units improving generalization. Experimental results on benchmark datasets show the good performance of our approach compared to previous methods.