USING MULTI-ANGLES EVOLUTIONARY ALGORITHMS FOR TRAINING TSK-TYPE NEURO-FUZZY NETWORKS

The development of a global-based method for building robust neuro-fuzzy networks has become an interesting issue. Among the various building methods, the evolutionary algorithms provide robust ways increasing the chances of meeting the optimal solution. However, evolutionary algorithms may only use a single angle to evaluate the searching space to obtain the optimal solutions. It implies that they may slowly or even hardly meet the optimal solution. Thus, the current study provides a novel architecture that uses multiple angles for evaluating the searching space. More specifically, the novel architecture adopts multiple angles to improve the evolutionary process by dynamically adjusting the searching space. By doing so, the proposed architecture can increase the chances of meeting the optimal solution. As shown in the results, the proposed architecture outperforms other existing evolutionary algorithms. Based on the results, a framework is proposed to build a benchmark for developing evolutionary algorithms that consider the multiple angles of the solution space.