Updating the Consequent Part of Fuzzy Rules in Parallel Distributed Fuzzy Genetics-Based Machine Learning

Abstract

In our former studies, we have proposed parallel distributed fuzzy genetics-based machine learning (GBML) for the speedup of fuzzy GBML. In our parallel distributed model, a population and training data are divided into subpopulations and training data subsets, respectively. A pair of a subpopulation and a training data subset is assigned to an island. One characteristic feature in our model is training data rotation over the islands. In this paper, we discuss how to update the consequent part of existing fuzzy rules after the training data rotation.