The differential evolution (DE) optimization algorithm predominantly relies on elite individuals and random difference to direct evolution. Although the strategy is clear and easy to implement, identifying a suitable direction for the DE mutation strongly depends on the direction information provided beforehand. To address this, we present a neighbor-induced mutation operator that simulates the neighbor-induced movement of Antarctic krill to guide the evolution direction in a natural manner. Additionally, center dispersion is proposed to disperse the population and redistribute individual positions to escape search stagnation, inspired by the spreading out of krill around newly discovered food. Comprising the new operator and the center dispersion pattern, this paper proposes a neighbor-induced DE algorithm with dispersion pattern (NDEd). The results of the comparative experiments verify the effectiveness of the neighbor-induced mutation operator and the dispersion pattern. Further, experimental results from 28 test functions of CEC2013 demonstrate that NDEd performs better compared to the other classic DE algorithms.

Introduction

In 1995, Storn and Price proposed the differential evolution (DE) algorithm [1], a practical, robust, and simple global optimization algorithm. Since then, DE and its variants have become some of the most competitive evolutionary computing algorithms. DE algorithms have been successfully applied in various scientific and engineering fields, such as mechanical engineering design [2], signal processing [3], chemical engineering [4], machine intelligence, and pattern recognition [5]. DE algorithms have exhibited outstanding performance when dealing with optimization problems. However, they have problems such as slow convergence speed and susceptibility to local optima [6]. Properly guiding the direction of evolution may be a key to solving these problems. For example, Cai and Wang [7] proposed a DE frame with neighborhood and direction information. However, the frame heavily depends on the selection of direction information. Although they combined an adaptive operator selection (AOS) from the available direction information with their algorithm in their subsequent research [8], selecting the most suitable type of direction information for the specific DE mutation strategy depends on the classification of the direction information given beforehand. Thus, it is difficult to implement automatic selection of the most suitable direction information in practice.