论文部分内容阅读
线性模型回归系数的一些稳健估计如LMS、LQS、LTS、LTA的应用越来越广泛,然而它们的精确计算依赖于NP难题,在遇到高维大规模数据集时不可能在较短时间内得到精确解.为尽快得到较高精度的近似解,提出了求解线性模型的稳健参数估计的整数编码遗传算法,通过计算机模拟试验验证了算法可以更快地找出全局最优解.
Some robust estimates of regression coefficients of linear models such as LMS, LQS, LTS and LTA are more and more widely used. However, their exact calculation depends on the NP problem and can not be solved in a short time in the case of high-dimensional large-scale data sets Get the exact solution.In order to get the approximate solution with high accuracy as soon as possible, an integer coding genetic algorithm for solving the robust parameter estimation of the linear model is proposed. The computer simulation shows that the algorithm can find the global optimal solution more quickly.