Fusion with distance-aware selection strategy for dandelion algorithm

摘要

The Dandelion Algorithm (DA) is a recently proposed swarm intelligent optimization algorithm which is inspired by the process of dandelion sowing. In DA, the selection strategy (SS) is of key importance which is responsible for choosing the dandelions to enter the next generation. However, current SS in DA only considers the function fitness value, which cannot maintain the diversity of the population and can easily fall into local optimal solutions prematurely. To address this problem, a distance-aware selection strategy (DSS) is proposed by jointly considering the function fitness value and the individual position distance. In addition, inspired by ensemble learning, a fusion based adaptive selection strategy (FSS) is proposed to further improve the performance of DA. Specifically, a fusion pool including several famous SSs and our proposed DSS is established. And in each iteration, the best SS in terms of overall reward is selected. Experimental results show that the proposed algorithms are powerful in terms of accuracy in the 28 standard functions from the CEC2013. For DSS, it outperforms several famous SSs in DA. For FSS, it is superior or competitive to 11 participating algorithms benchmarked on CEC2013 and five recently proposed algorithms improved by new selection strategy. Since FSS is a universal framework, it can be applied to other swarm intelligence algorithms with selection strategy.