摘要
分析目前基于聚类思想的遗传算法的不足,提出一种基于生长树聚类的改进型遗传算法。采用最小生成树的聚类方法,能对形状复杂且非重叠样本的候选解进行聚类形成家族;新的族间交叉算子保持了种群的多样性;改进的族内交叉算子和改进的变异算子使得算法在后期仍能快速收敛;实验对经典算法测试函数进行优化,并与其他算法的优化结果对比,从而说明改进型遗传算法的性能。实验结果表明:基于生长树聚类的改进型遗传算法能有效提高求解精度,快速搜索到最优解。
The shortcomings of present genetic algorithm based on clustering thoughts are analyzed,and a new advanced genetic algorithm based on propagating tree clustering is proposed.It uses clustering method of minimum spanning tree and can cluster candidate solutions of non-overlap samples in complex shape and generate new families;the new inter-family crossover operators maintain population＇s multiplicity,the improved intra-family crossover operator and mutation operator can make the algorithm keep rapid convergence in later phase.The experiment optimized several classical algorithm trial functions,and compared them with other algorithms＇ optimized results to demonstrate the performance of the advanced genetic algorithm.The test results indicated that the advanced genetic algorithm based on propagating tree clustering can increase solution＇s precision effectively and search optimal solution quickly.
出处
《计算机应用与软件》
CSCD
2010年第1期127-130,共4页
Computer Applications and Software
基金
黑龙江省自然科学基金(F200605)
黑龙江省教育厅海外学人合作项目(1153h21)
关键词
遗传算法
生长树
聚类
族间交叉
Genetic algorithm Propagating tree Clustering Inter-family crossover
