ART and ARTMAP neural networks for applications: Self-organizing learning, recognition, and prediction

Author(s): Carpenter, G.A. | Gjaja, N.M. | Gopal, S. | Markuzon, N. | Woodcock, C.E. |

Year: 1997

Citation: In L.C. Jain (Ed.), Soft Computing Techniques in Knowledge-Based Intelligent Systems in Engineering, New York: Springer-Verlag, 279-317.

Abstract: ART and ARTMAP neural networks for adaptive recognition and prediction have been applied to a variety of problems. Applications include parts design retrieval at the Boeing Company, automatic mapping from remote sensing satellite measurements, medical database prediction, and robot vision. this chapter features a self-contained introduction to ART and ARTMAP dynamics and a complete algorithm for applications. Computational properties of these networks are illustrated by means of remote sensing and medical database examples. The basic ART and ARTMAP networks feature winner-take-all (WTA) competitive coding, which groups inputs into discrete recognition categories. WTA coding in these networks enables fast learning, that allows the network to encode important rare cases but that may lead to inefficient category proliferation with noisy training inputs. This problem is partially solved by ART-EMAP, which uses WTA coding for learning but distributed category representations for test-set prediction. In medical database prediction problems, which often feature inconsistent training input predictions, the ARTMAP-IC network further improves ARTMAP performance with distributed prediction, category instance counting, and a new search algorithm. A recently developed family of ART models (dART and dARTMAP) retains stable coding, recognition, and prediction, but allows arbitrarily distributed category representation during learning as well as performance.

Topics: Machine Learning, Models: ART 1, ARTMAP, Distributed ART,

PDF download

Cross References

  1. Frey & Slate Letter data
    This data set describes statistical attributes of 20,000 digitized pictures of letters, and was used to study machine learning using Holland-style adaptive classifiers (Frey & Slate, 1991). Our copy was obtained from the UCI ... Data Details