Transfer-Based Particle Swarm Optimization for Large-Scale Dynamic Optimization With Changing Variable Interactions
提出一种基于迁移的粒子群优化算法，通过动态差分分组在线分解变量、利用解迁移策略应对环境变化，在高达1000维的问题上优于现有算法。
Cooperative coevolutionary algorithms are popular to solve large-scale dynamic optimization problems via divide-and-conquer mechanisms. Their performance depends on how decision variables are grouped and how changing optima are tracked. However, existing decomposition methods are computationally expensive, resulting in limitations under dynamic variable interactions. Quick online decomposition is still a challenging issue, along with solution reconstruction for new subproblems. This paper proposes transfer-based particle swarm optimization, which adopts a dynamic differential grouping for online decomposition and a solution transfer strategy in response to environmental changes. Particularly, once an environmental change occurs, the dynamic differential grouping readjusts historical groupings based on the change severity of variable interactions. In addition, according to the similarity between subproblems in successive environments, the solution transfer strategy constructs new solutions from historical ones through dimension mapping. Multiple swarms are created to explore subareas of subproblems. Experimental results show that the proposed algorithm outperforms state-of-the-art algorithms on problem instances up to 1000-D in terms of solution optimality. The dynamic differential grouping obtains accurate groupings using less function evaluations.