A Collection of Art-Family Graphical Simulations
MetadataShow full item record
The Adaptive Resonance Theory (ART) architecture, first proposed by (Grossberg, 1976b, 1976a), is a self-organizing neural network for stable pattern categorization in response to arbitrary input sequences. Since its original formulation, several versions of ART have been proposed, each designed to handle a particular task or input format. Recent ART architectures have been designed to work in a supervised fashion, offering a viable alternative to supervised neural networks such as backpropagation (Rumelhart, Hinton, & Williams, 1986). Perhaps the best-known variant of ART is ART2 (Carpenter & Grossberg, 1987b), an unsupervised neural network that handles analog inputs. We have developed a series of simulators for some of the ART-family neural architectures, namely, ART2 (Carpenter & Grossberg, 1987b), ART2-A (Carpenter, Grossberg, & Rosen, 1991b), Fuzzy ART (Carpenter, Grossberg, & Rosen, 1990), and Fuzzy ARTMAP (Carpenter, Grossberg, Markuzon, & Reynolds, 1992). This article briefly summarizes the history and functionality of ART and its variants, and then describes the software package, which is available in the public domain.
RightsCopyright 1994 Boston University. Permission to copy without fee all or part of this material is granted provided that: 1. The copies are not made or distributed for direct commercial advantage; 2. the report title, author, document number, and release date appear, and notice is given that copying is by permission of BOSTON UNIVERSITY TRUSTEES. To copy otherwise, or to republish, requires a fee and / or special permission.