INVERTING GENERATING FUNCTIONS WITH INCREASED NUMERICAL PRECISION-A COMPUTATIONAL EXPERIENCE  

作　　者：Nam K. KIM Mohan L. CHAUDHRY Bong K. YOON Kilhwan KIM 

机构地区：[1]Department of Industrial Engineering, Chonnam National University [2]Department of Mathematics and Computer Science, Royal Military College of Canada [3]Defense Management College, Korea National Defense University [4]Internet Research Lab., Electronics and Telecommunications Research Institute

出　　处：《Journal of Systems Science and Systems Engineering》2011年第4期475-494,共20页系统科学与系统工程学报（英文版）

基　　金：support provided by the Department of Mathematics and Computer Science, Royal Military College of Canada, Kingston where he held a Visiting Research Fellowship under NSERC Grant

摘　　要：In this paper, we consider the numerical inversion of a variety of generating functions （GFs） that arise in the area of engineering and non-engineering fields. Three classes of GFs are taken into account in a comprehensive manner： classes of probability generating functions （PGFs） that are given in rational and non-rational forms, and a class of GFs that are not PGFs. Among others, those PGFs that are not explicitly given but contain a number of unknowns are largely considered as they are often encountered in many interesting applied problems. For the numerical inversion of GFs, we use the methods of the discrete （fast） Fourier transform and the Taylor series expansion. Through these methods, we show that it is remarkably easy to obtain the desired sequence to any given accuracy, so long as enough numerical precision is used in computations. Since high precision is readily available in current software packages and programming languages, one can now lift, with little effort, the so-called Laplacian curtain that veils the sequence of interest. To demonstrate, we take a series of representative examples： the PGF of the number of customers in the discrete-time Geo^X/Geo/c queue, the same in the continuous-time M^X/D/c queue, and the GFs arising in the discrete-time renewal process.In this paper, we consider the numerical inversion of a variety of generating functions （GFs） that arise in the area of engineering and non-engineering fields. Three classes of GFs are taken into account in a comprehensive manner： classes of probability generating functions （PGFs） that are given in rational and non-rational forms, and a class of GFs that are not PGFs. Among others, those PGFs that are not explicitly given but contain a number of unknowns are largely considered as they are often encountered in many interesting applied problems. For the numerical inversion of GFs, we use the methods of the discrete （fast） Fourier transform and the Taylor series expansion. Through these methods, we show that it is remarkably easy to obtain the desired sequence to any given accuracy, so long as enough numerical precision is used in computations. Since high precision is readily available in current software packages and programming languages, one can now lift, with little effort, the so-called Laplacian curtain that veils the sequence of interest. To demonstrate, we take a series of representative examples： the PGF of the number of customers in the discrete-time Geo^X/Geo/c queue, the same in the continuous-time M^X/D/c queue, and the GFs arising in the discrete-time renewal process.

关 键 词：QUEUING applied probability numerical inversion generating function 

分 类 号：O213[理学—概率论与数理统计]