摘要
在动态测试数据处理中,常常要进行稳健回归分析和最小最大值回归分析。讨论了遗传算法及其在lp数据拟合中的应用。由于遗传算法是用点群进行寻优,而不是用一个单点进行寻优,具有隐含并行算法的特点;群体在每一代的进化过程中执行同样的复制、交叉、变异操作,仅使用问题本身所对应的适应度函数,而不需要任何其它先决条件或辅助信息;遗传算法使用随机转换规则,而不是确定性规则进行运算。从而使得遗传算法是一类全局收敛算法,能够在lp数据拟合中得到很好的应用。给出了应用实例。
Robust regression analysis and minimax residual error analysis are two aspects in data processing of dynamic measurement. The genetic algorithms and its application on lp data fitting is described. Genetic algorithms model natural processes, such as selection, recombination, mutation and search a population of points in parallel, is not just a single point. No derivative information or other auxiliary knowledge but the objective function and corresponding fitness are required to get the directions of search and probabilistic transition rules, not deterministic ones are used in the algorithms. Correct results of lp data fitting can be gotten by using genetic algorithms which bear global convergence. Examples and related results are also given.
出处
《计量学报》
EI
CSCD
北大核心
2005年第3期284-288,共5页
Acta Metrologica Sinica
