The Learning Algorithm for a Novel Fuzzy Neural Network
Symmetric polygonal fuzzy numbers are employed to construct a class of novel feedforward fuzzy neural networks (FNN’s)—the polygonal FNN’s. Their input–output (I/O) relationships are built upon a novel fuzzy arithmetic and extension principle for the polygonal fuzzy numbers. We build the topological architecture of a three layer polygonal FNN, and present its I/O relationship representation. Also the fuzzy BP learning algorithm for the polygonal fuzzy number connection weights and thresholds is developed based on calculus of max–min (∨– ∧) functions. At last some simulation examples are compared to show that our model possess strong I/O ability and generalization capability.
Unable to display preview. Download preview PDF.
- 6.Buckley, J.J., Hayashi, Y.: Direct Fuzzification of Neural Networks. In: Proc. of 1st Asian Fuzzy Sys. Symp., Singapore, vol. 1, pp. 560–567 (1993)Google Scholar
- 22.Liu, P., Li, H.: Symmetric Polygonal Fuzzy Numbers. In: The 9th International Conference on Fuzzy Theory and Technology, NC, USA, pp. 26–30 (2003)Google Scholar