基于演化搜索信息的量子行为粒子群优化算法

Improved QPSO algorithm based on search history

针对量子行为粒子群优化算法可能过早收敛而陷入局部最优的问题,提出了基于演化搜索信息的非重复访问量子行为粒子群优化算法(Non-revisited QPSO,NrQPSO)。该算法将演化搜索信息记录方案和标准QPSO算法结合起来,确保所有更新的粒子位置都是未被重复访问的,并通过变异操作增加粒子的多样性。演化搜索信息记录方案利用二维空间分割树(BSP)将连续搜索空间划分为不同的重叠子区域,并且将子区域作为粒子变异范围,使得相应的变异操作是一种无参数的自适应变异。对比其他传统算法,通过对八个标准测试函数的实验结果表明,NrQPSO算法在处理多峰和单峰测试函数时具有更好的优化性能,收敛精度和收敛速度都得到了提高,证明该算法的有效性。

An improved Non-revisited QPSO algorithm based on search history(NrQPSO)is proposed to help prevent premature convergence and stagnate at local optimal solutions.NrQPSO is an integration of the entire search history scheme and a standard Quantum-behaved Particle Swarm Optimization(QPSO).It guarantees that all updated positions are not revisited before and the diversity of particles is increased by mutation.The search history scheme partitions the continuous search space into overlap sub-regions by using BSP tree.The partitioned sub-region serves as mutation range such that the corresponding mutation is adaptive and parameterless.Compared with other traditional algorithms,the experiment results on eight standard testing functions show that the proposed algorithm is superior regarding the optimization of multimodaland unimodal functions,with enhancement in both convergence speed and precision.Those demonstrate the effectiveness of the algorithm.