A mixed mutation approach for evolutionary programming based on guided selection strategy

Abstract

A proper balance between exploration and exploitation is essential to maintain adequate genetic diversity within the evolving population of an evolutionary algorithm (EA). Early loss of genetic diversity causes premature trapping around the locally optimal points of the fitness landscape. Evolutionary programming (EP), one of the major branches of EA, obtains exploration and exploitation abilities by mutation operators. As one single mutation operator is not sufficient, mixing several explorative and exploitative mutation operators can improve the performance of EP. This paper presents a mixed mutation scheme for EP based on a guided selection strategy. This strategy guides the participation of mutation operators throughout the evolutionary process. The proposed algorithm has been examined on a test-suite of 20 benchmark functions. Experimental results show that combining different mutation operators along with the guided selection strategy significantly enhance the performance of EP.