SOME PROPERTIES ON FUZZY INFERENCE SYSTEMS COMPOSED OF SMALL NUMBER OF INPUT RULE MODULES
The automatic construction of fuzzy system with a large number of input variables involves many difficulties such as large time complexity and getting stuck in a shallow and local minimum. As models to overcome them, the SIRMs (single input rule modules) and DIRMs (double input rule modules) models have been proposed. In some numerical simulations such as EX-OR and control problems, it has been shown that there exists the difference of the capability between DIRMs and SIRMs models. In this paper, we theoretically show the difference of the capability among SNIRMs (small number of input rule modules) models. As a result, there exists the difference of the capability among SNIRMs models. Further, in order to reduce the number of modules of the SNIRMs model effectively, we propose two types of learning algorithms. The one, called pruning algorithm for learning, deletes some rule modules in the SNIRMs model. The other one, called generative algorithm for learning, incrementally adds some modules to the SNIRMs model. The simulation result shows that the SNIRMs models obtained by the proposed algorithms are superior in terms of accuracy compared to the conventional SIRMs model and in terms of the number of modules to the conventional SNIRMs model.
fuzzy inference system, SIRMs model, SNIRMs model, generative algorithm, pruning algorithm.