摘要
设函数系{φ_j(x)}_(j=1)~n 在离散点集 X 上线性无关.在这个条件下,[10]研究了最佳 L_1逼近解集的几何性质与特征.本文给出了一个新的下降算法,并证明了它具有有限步收敛的性质.通过实际上机运行表明,它比目前国际上现行的 BR 算法,BCS 算法及 BS 算法更为节省 CPU 时间.
The geometric properties and characters for best discrete L_1 Approxi- mation have been studied in〔10〕under condition that the system of functions {(?)j}~n_j=1 is independent in a discrete point set X.Under same condition,a new descent algorithm is given and its convergence in finite steps is proved.Through practice tuning on computer,new algorithm is the most efficient than other current algorithms.
关键词
最佳逼近
函数逼近论
离散L1逼近
best approximation
approximation theory of functions/discrete L_1 approximation
