【摘要】 对于离散的多面体范数逼近问题提供了两个带有二阶校正的算法,它们无需显著地增加程序的复杂性和计算量,但可以在一阶线性近似招致较大误差时提高算法的有效性。本文还论证了算法的收敛性,并对算法的实现作了一些讨论。更多还原

【Abstract】 For the discrete polyhedral norm approximation problems, which includes the very useful 1l norm and 1∞ norm approximation problems as its special cases, two algorithms are proposed in this paper, both using some second order correction technique. Without remarkably increasing the complexity and the computational effort in each iteration than those of the method in [3], these algorithms may enhance the effectiveness of the previous method when first order approximation incurs big error. Convergence of both algorithms is proved and their implementation is also discussed briefly.更多还原