Nonlinear system modeling by competitive learning and adaptivefuzzy inference system

摘要

Modeling nonlinear systems by neural networks and fuzzy systemsnencounters problems such as the conflict between overfitting and goodngeneralization and low reliability, which requires a great number ofnfuzzy rules or neural nodes and uses very complicated learningnalgorithms. A new adaptive fuzzy inference system, combined with anlearning algorithm, is proposed to cope with these problems. First, thenalgorithm partitions the input space into some local regions byncompetitive learning, then it determines the decision boundaries fornlocal input regions, and finally, based on the decision boundaries, itnlearns the fuzzy rule for each local region by recursive least squaresn(RLS). In the learning algorithm, the key role of the decisionnboundaries is highly emphasized. To demonstrate the validity of thenproposed learning approach and the new adaptive fuzzy inference system,nfour examples are studied by the proposed method and compared with thenprevious results