数値微分における最適刻み間隔の決め方とその応用

概要

論文の詳細を見る
Differentiation is one of the useful operation in the information processing, but its numerical calculation may lead to noticeable error. In this paper mean square error, which arises from transforming a tabulated function into a derived function in a certain closed interval, is taken into consideration. The truncation error is expressed in quadratic form on the coefficients of the Kth-degree polynomial with which the tabulated function is approximated. Using this expression, the optimum interval for numerical differentiation which minimizes the total error, namely the truncation error and the round-off error, is determined. As the application of the determining method of the optimum interval, the problem to determine the class internal for frequency distribution is discussed.
一般社団法人情報処理学会の論文
1978-02-15