Fuzzy Inference Network with Mamdani Fuzzy Inference System

Abstract

In the modern era, the amount of data generated is increasing at an exponential rate. The generated data has both numeric as well as linguistic form. Learning or extracting relevant information from these types of data is a major challenge for researchers. In this chapter, we have proposed a generic architecture of a network built from Mamdani fuzzy inference system as its basic building blocks and it tries to learn the information from data. Each node of the network acts as a complete Mamdani fuzzy inference system mapping numeric as well as linguistic information of the data from input to output in terms of linguistic rule-based inference. Parameters of the input fuzzy membership functions appearing in the premise parts and output fuzzy membership functions appearing in consequent parts of the rules in the fuzzy rule base of each node in the network constitute overall parameters of the network. The proposed model is trained using advanced optimization techniques to optimize the network parameters for better performance. The effectiveness of the trained model is tested on two different datasets. The proposed approach is compared with the Takagi-Sugeno Fuzzy Inference Network and feed-forward artificial neural network with similar architecture.