In this paper, a new variant of the Radial Basis Function Network with the Dynamic Decay Adjustment algorithm (i.e., RBFNDDA) to undertake data classification problems is proposed. The new network is formed by integrating the learning algorithm of the Fuzzy ARTMAP (FAM) neural network into RBFNDDA. The proposed RBFNDDA-FAM network inherits the salient features of FAM and overcomes the shortcomings of the original RBFNDDA network. The effectiveness of RBFNDDA-FAM is demonstrated using two benchmark problems. The first involves an artificial data set whereas the second uses a medical data set related to thyroid diagnosis. The results from these studies are compared, analyzed, and discussed. The outcomes positively reveal the potentials of RBFNDDA-FAM in learning information with a compact network architecture, in addition to high classification performances.