Enhanced interactive differential evolution using evolutionary level

Differential evolution, one of the evolutionary algorithms, is a population-based stochastic search technique for solving optimization problems in a continuous space. Due to its simplicity, effectiveness and robustness, DE has been applied to a variety of real-world problems. As one of the successful application, DE is incorporated to interactive evolutionary computation (IEC) framework. Interactive DE (DDE) which enables the optimization that considers subjective preference and sensitivity were reported. However, human fatigue is one of the most important problems of IDE. In this paper, we aim at development of fast method of IDE using evolutionary level for reducing human fatigue. The evolutionary level is an indicator that corresponds to the number of generation alterations. In the proposed method, a high level individual has a higher chance to be selected as a parent by introducing roulette wheel selection using evolutionary level. Through the experimental results, we confirm that the convergence speed of the proposed method is faster than conventional DE.