This paper presents a new approach to incrementally constructing a neural network that is capable of learning new information without forgetting old knowledge. The proposed neural network, called hyper-spherical ARTMAP network (HS-ARTMAP network), is a sy
