COMPARING EVOLUTIONARY ALGORITHMS ON TUNING THE PARAMETERS OF FUZZY WAVELET NEURAL NETWORK

Document Type: Research Paper

10.22099/ijste.2013.1888

Abstract

In recent years Fuzzy Wavelet Neural Networks (FWNNs) have been used in many
areas. Function approximation is an important application of FWNNs. One of the main problems
in effective usage of FWNN is tuning of its parameters. In this paper several different evolutionary
algorithms including Genetic Algorithm (GA), Gravitational Search Algorithm (GSA),
Evolutionary Strategy (ES), Fast Evolutionary Strategy (FES) and variants of Differential
Evolutionary algorithms (DE) are used for adjusting these parameters on five test functions. The
obtained results are compared based on some measures by using multiple non-parametric
statistical tests. The comparison reveals the superiority of some variants of DE in terms of
convergence behavior and the ability of function approximation.

Keywords