Evolutionary programming with non-coding segments for real-valued function optimization

Evolutionary programming (EP) constitutes a class of general optimization algorithms based on the model of natural evolution. Real-valued function optimization has also been included in this domain since the early 90's. One of the main features is that the mutation size is self-adaptively controlled by additional parameters (strategy parameters). However, it was found that the mutation size control is not "sufficiently" self-adaptive: EP often practically stops its genetic search before reaching a global optimum, due to the tendency that strategy parameters converge in the early generations, with the result that the population cannot move to another place. In the paper, in order to overcome this burdensome character, an extended EP that utilizes the genetic drift as another source of changing strategy parameters is proposed. Computer simulations were conducted using several test functions in order to evaluate the performance of the proposed method.