A Fuzzy Rule Structured Neural Network(Journal of Japan Society for Fuzzy Theory and Systems)

概要

論文の詳細を見る
This paper proposes a fuzzy rule structured neural network, which is an approximate reasoning system constructed by neural network components. This system accomplish approximate reasoning by the interpolating capabilities of neural networks. Each neural network component executes an elementary function which is a part of an approximate reasoning procedure. The initial state of membership functions in the antecedent part are based on a priori knowledge, and the initial system categorizes practical data to be stored as the conclusion part. The membership functions in the antecedent part are modified by using solutions of the inverse problem of a neural network. An application to a system identification problem clarifies the efficiency of the proposed system.
日本知能情報ファジィ学会の論文
1992-10-15