Experimental investigation with analyzing the training method complexity of neuro-fuzzy networks based on parallel random search

Abstract

The problem of training neuro-fuzzy networks is discussed. The computational complexity of the method for training neuro-fuzzy networks on the basis of parallel random search is analyzed. Theoretical estimations of the speedup and efficiency of the method are found. Software implementing of the method in C++ with using the MPI library and providing the construction of neuro-fuzzy networks in terms of the given observation sets is developed. Experiments for practical tasks are carried out.