A Neural Network Equivalent to a Neuro-Fuzzy System for Classification

摘要

The paper presents a neural network that is equivalent to a connectionist neuro-fuzzy system designed for classification problems. The neural network is a single-layer network that contains classical neurons with a specific activation function. The weights of the linear part of the neurons as well as the nonlinear functions have an interpretation related to the corresponding neuro-fuzzy system. Examples of classification problems solved by means of this network are illustrated.