A preliminary study on mutation operators in cooperative competitive algorithms for RBFN design

Evolutionary Computation is a typical paradigm for the Radial Basis Function Network design. In this environment an individual represents a whole network. An alternative is to use cooperative-competitive methods where an individual is a part of the solution. CO²RBFN is an evolutionary cooperative-competitive hybrid methodology for the design of Radial Basis Function Networks. In the proposed cooperative-competitive environment, each individual represents a Radial Basis Function, and the entire population is responsible for the final solution. In order to calculate the application probability of the evolutive operators over a certain Radial Basis Function, a Fuzzy Rule Based System has been used. In this paper, CO²RBFN is adapted to the regression problem and an analysis of mutation operator is performed. To do so, two implementation of the mutation operator, based on gradient and based on clustering, have been implemented and tested. The results have been compared with other data mining and mathematical methods usually used in regression problems.