Identification of Dynamic Plants Using Fuzzy Wavelet Network: A Multimodal Memetic Approach

Fuzzy Wavelet Neural Network (FWNN) is a subject in computer science which integrates fuzzy logic, neural network, and wavelet functions to achieve an admissible modeling tool. Training is the most important issue in FWNN. Many training methods have been proposed in the literature. However, there is no trying to train FWNN with multimodal optimization. It is desirable to obtain multiple global and local optima in a single run. In this paper, a multimodal memetic algorithm is proposed for training FWNN. A Particle Swarm Optimization (PSO) using ring neighborhood topology is employed to maintain diversity in different niches. Local searches are active during PSO running to improve the exploitation ability of the proposed method. Experimental results over two system identification problems show the superior performance of the proposed multimodal memetic algorithm.