Self-adaptive mutation strategy for evolutionary programming based on fitness tracking scheme

In order to achieve a satisfactory optimization performance by evolutionary programming (EP), it is necessary to ensure proper balance between global exploration and local exploitation. It is obvious that one single mutation operator is not the answer. Moreover, early loss of genetic diversity causes premature trapping around locally optimal points of the fitness landscape. This paper presents a fitness tracking based evolutionary programming (FTEP) algorithm incorporating a fitness tracking scheme to find the locally trapped individuals and treat them in a different way so that they are able to improve their performance. In comparison with other EP based algorithms, FTEP incorporates several mutation operators in one algorithm and employs a self-adaptive strategy to gradually self-adapt the mutation operators in order to apply an appropriate mutation operator on the individual based on its need. A test-suite of 25 benchmark functions has been used to evaluate the performance and results have been compared with some recent evolutionary systems. The experimental results show that FTEP often performs better than most other algorithms on most of the problems.