A fuzzy min-max neural network with classification performance irrelevant to the input sequences of samples
Jing Hu, Yiyuan Luo · 2016
This paper presented a method for solving the dependence of classification performance on the input sequences of samples in learning algorithm of the fuzzy min-max neural network proposed by Simpson through calculating the values of similarity matrix. To realize the aim, the serial input of samples was changed to parallel input, so as to highlight the integral structure of all the samples instead of the internal structure of current samples in the membership function. The experimental results showed that the new learning method in this paper was feasible and effective.