A Novel Modified Tree-Seed Algorithm for High-Dimensional Optimization Problems

• Published in: Chinese Journal of Electronics Vol. 29; no. 2; pp. 337 - 343
• Main Authors: Zhao, Shijie, Gao, Leifu, Tu, Jun, Yu, Dongmei
• Format: Journal Article
• Language: English
• Published: Published by the IET on behalf of the CIE 01.03.2020
• Subjects: adjacent‐generation best trees; control parameter; convergence; convergence efficiency; global exploration; gradient information; gradient methods; High‐dimension optimization; high‐dimensional continuous optimization problems; high‐dimensional optimization problems; iterative methods; local exploitation process; local optima avoidance; Local reinforcement; local reinforcement strategy; LRS strategy; MTSA; novel modified tree‐seed algorithm; offspring seeds; optimisation; probabilities; probability; SDSC; seed domain shrinkable coefficient; Seed domain shrinkage; trees (mathematics); Tree‐seed algorithm; iterative methods; convergence; gradient information; SDSC; novel modified tree-seed algorithm; local exploitation process; LRS strategy; MTSA; optimisation; seed domain shrinkable coefficient; control parameter; Local reinforcement; gradient methods; probabilities; high-dimensional continuous optimization problems; high-dimensional optimization problems; probability; trees (mathematics); local optima avoidance; Seed domain shrinkage; offspring seeds; Tree-seed algorithm; local reinforcement strategy; adjacent-generation best trees; convergence efficiency; global exploration; High-dimension optimization
• ISSN: 1022-4653; 2075-5597; 2075-5597
• DOI: 10.1049/cje.2020.01.012

To efficiently handle high-dimensional continuous optimization problems, a Modified tree-seed algorithm(MTSA) is proposed by coupling a newly introduced control parameter named as Seed domain shrinkable coefficient(SDSC) and Local reinforcement strategy(LRS) based on gradient information of adjacentgeneration best trees. SDSC is dynamically decreased with iterations to adjust the produced domain of offspring seeds, for achieving the tradeoff between the global exploration and local exploitation. LRS strategy is to execute local exploitation process by employing gradient information of two best trees, for enhancing convergence efficiency and local optima avoidance with probabilities. The compared experimental results show the different effects of differenttype SDSC on MTSA, the faster convergence efficiency and the stronger robustness of the proposed MTSA.