Fuzzy LAPART Supervised Learning Through Inferencing for Stable Category Recognition

Author(s): Fausett, L.V. | Ham, F.H. | Han, G. |

Year: 1995

Citation: Journal of Artificial Neural Networks, Vol. 2, No. 3.

Abstract: Fuzzy LAPART (laterally primed adaptive resonance theory), a neural network architecture for supervised learning through logical inferencing, is introduced with fast and slow learning algorithms and match tracking capability. Based on the original architecture developed by Healy, et al., the enhanced architecture consists of interconnected fuzzy adaptive resonance theory (fuzzy ART) modules originated by Carpenter, et al. The interconnections enable fuzzy LAPART to infer one pattern class from another to form a predictive pattern class. Slow learning capability has been incorporated into the neural network with fast commit and slow recode options. The problem of separation of spirals is used to perform benchmark tests for fuzzy LAPART. Also, based on fuzzy set theory, geometric interpretations are presented in 2 and 3 dimensional spaces using fuzzy LAPART. Performance results for both test cases are compared to results obtained from a counterpropagation clustering network

Topics: Machine Learning, Models: ART 2 / Fuzzy ART, Modified ART,

PDF download

Cross References