Training Fuzzy Number Neural Networks with Alpha-cut Refinements
In a fuzzy number neural network, the inputs, weights, and outputs are general fuzzy numbers. The requirement that F¯α(1) ⊂F¯α(2) whenever α(1)>α(2) imposes an enormous number of constraints on the weight parameterizations during training. This problem can be solved through a careful choice of weight representation. This new representation is unconstrained, so that standard neural network training techniques may be applied. Unfortunately, fuzzy number neural networks still have many parameters to pick during training, since each weight is represented by a vector. Thus moderate to large fuzzy number neural networks suffer from the usual maladies of very large neural networks. In this paper, we discuss a method for effectively reducing the dimensionality of networks during training. Each fuzzy number weight is represented by the endpoints of its α-cuts for some discretization 0⩽α1<α2<...<αn ⩽1. To reduce dimensionality, training is first done using only a small subset of the αi. After successful training, linear interpolation is used to estimate additional α-cut endpoints. The network is then retrained to tune these interpolated values. This refinement is repeated as needed until the network is fully trained at the desired discretization in &alpha.
J. P. Dunyak and D. C. Wunsch, "Training Fuzzy Number Neural Networks with Alpha-cut Refinements," Systems, Man, and Cybernetics, 1997. IEEE International Conference on Computational Cybernetics and Simulation, vol. 1, pp. 189-194, Institute of Electrical and Electronics Engineers (IEEE), Jan 1997.
The definitive version is available at https://doi.org/10.1109/ICSMC.1997.625747
IEEE International Conference on Computational Cybernetics and Sumulation: Systems, Man and Cybernetics (1997: Oct. 12-15, Orlando, FL)
Electrical and Computer Engineering
International Standard Book Number (ISBN)
International Standard Serial Number (ISSN)
Article - Conference proceedings
© 1997 Institute of Electrical and Electronics Engineers (IEEE), All rights reserved.
01 Jan 1997