An Efficient Mapping of Fuzzy Art Onto a Neural Architecture
It is possible to eliminate the reset/resonance cycle of many of the ART algorithms without modifying their learning or classification rules. The enabling principle is to check the vigilance criterion before comparing the choice functions, rather than vice versa. This re-ordering of operations reduces the computational complexity. It leads to a simpler neural architecture that is better suited to parallel implementation and facilitates understanding the algorithm.
KeywordsInput Vector Choice Function Match Criterion Efficient Mapping Neural Architecture
Unable to display preview. Download preview PDF.
- Blume, M. and Esener, S.C. (1995), “Optoelectronic fuzzy ARTMAP processor, ” 1995 OSA Technical Digest Series, vol. 10, pp. 213–215.Google Scholar
- Serrano-Gotarredona, T., Linares-Barranco, B., and Huertas, J.L. (1995), “A real time clustering CMOS neural engine, ” Advances in Neural Information Processing Systems, vol. 7, pp. 755–760.Google Scholar
- Carpenter, G.A., Grossberg, S., and Rosen, D.B. (1991), A Neural Network Realization of Fuzzy ART (Technical Report CAS/CNS-91–021), Boston University, Center for Adaptive Systems, Boston, MA.Google Scholar
- Rubin, M.A. (1995), “Issues in automatic target recognition from radar range profiles using fuzzy ARTMAP, ” Proceedings of the 1995 World Congress on Neural Networks, pp. 197–202.Google Scholar