Success-History Based Position Adaptation in Gaining-Sharing Knowledge Based Algorithm

Abstract

This paper introduces a modification of the recently developed Adaptive Gaining Sharing Knowledge (AGSK) algorithm. The AGSK algorithm simulates the process of human gaining and sharing knowledge using two main phases to solve optimization problems: junior and senior. AGSK’s efficiency was proved; however, there are still various approaches that can be used to improve its workability. In this study a new technique for generating potential solutions for the AGSK algorithm is proposed. This technique uses a historical memory of successful positions found by individuals stored in the external archive to guide those individuals in different directions and thus to improve the exploration and exploitation abilities of the AGSK algorithm. At first, the size of the external archive was fixed, but later in order to improve the performance of AGSK, a reduction technique was applied to decrease its size during the optimization process. Moreover, three different approaches were used to update the external archive after each algorithm’s iteration. The modified algorithm (as well as its original variant) was evaluated on a set of test functions taken from the CEC 2021 competition. The obtained experimental results are presented and compared. It was established that the proposed modification of the AGSK algorithm allows finding better solutions with the same computational effort. Thus, proposed position adaptation technique’s usefulness was demonstrated.