论文部分内容阅读
基于记忆的人工蜂群算法(ABCM)通过记住成功使用的邻居和系数指导人工蜂群下一步的搜索,需消耗多次函数评价收敛到吸引子,且始终使用与上次相同的排斥系数,造成收敛速度不快、多样性不足,易陷入局部最优解.提出一种改进算法,当使用吸引系数时,候选解只消耗一次函数评价收敛到吸引子,如果候选解好于当前解,则替换当前解,否则直接删除该记忆,这样可以利用尽量小的代价得到尽量大的收益.当使用排斥系数时,该系数的数值部分重新随机生成,以增加多样性和随机性,有利于算法跳出局部最优解.在22个不同类型函数上的实验表明,改进算法在收敛速度和精度方面明显优于人工蜂群算法和ABCM.
The memory-based artificial bee colony algorithm (ABCM), which guides the next search of artificial swarms by remembering the neighbors and coefficients that are successfully used, consumes multiple function evaluations to converge to attractors and always uses the same rejection coefficient as last time, Resulting in unpleasant convergence rate and lack of diversity, and easily fall into the local optimal solution.An improved algorithm is proposed, in which the candidate solution consumes only one function evaluation converges to the attractor when the attractive coefficient is used, and if the candidate solution is better than the current solution, The current solution, or directly delete the memory, so you can make use of as little as possible to get the maximum benefit.When using the exclusion coefficient, the numerical part of the coefficient re-generated randomly to increase the diversity and randomness, is conducive to the algorithm out of the local The optimal solution.The experiments on 22 different types of functions show that the improved algorithm is superior to artificial bee colony algorithm and ABCM in terms of convergence speed and accuracy.