Enhancing Tree-Seed Algorithm Via Feed-Back Mechanism for Optimizing Continuous Problems.

Tree-Seed Algorithm (TSA) is a novel population-based random search algorithm with its advantages in continuous optimization problems. However, there are some problems in its searching procedure. Problem (1): its balance mechanism of exploration and exploitation is implemented with a constant ST, and this fixed value is unreasonable in the random search procedure; Problem (2): the seed generation mechanism is achieved randomly without considering different searching phases based on function evaluations. To overcome these two problems, the feedback mechanism should be enhanced. Firstly, the st_TSA is proposed to solve the Problem (1); secondly, the ns_TSA is proposed to further solve the Problem (2); finally, in order to inherit these feedback mechanisms, a novel fb_TSA has been proposed and verified by standard 30 test benchmark functions from IEEE CEC 2014 with the basic TSA and its variants, such as STSA. In addition, GWO, ABC, SCA, DE, PSO and CLPSO are adopted for some comparative experiments with different dimensions. The computational results demonstrate that the enhanced feedback mechanism on ST and ns parameters can improve the optimization capability of the basic TSA significantly, especially in global optimum. The applicability of the proposed fb_TSA is proved by the 4 real engineering problems when compared with TSA, SCA, ABC and PSO.