Bounding Taylor approximation errors for the exponential function in the presence of a power weight function

A.J.E.M. Janssen

2024-03-04

Abstract:Motivated by the needs in the theory of large deviations and in the theory of Lundberg's equation with heavy-tailed distribution functions, we study for $n=0,1,...$ the maximization of

Classical Analysis and ODEs

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to estimate the bounds of the Taylor approximation error of the exponential function in the presence of power - function weights. Specifically, the researchers are concerned with how to maximize the following two expressions:

1. For \(n = 0, 1,\ldots\),
\[
S: \frac{1 - e^{-s}\left(1+\frac{s}{1!}+\ldots+\frac{s^{n}}{n!}\right)}{s^{\delta}}=E_{n,\delta}(s)\quad\text{for}\quad s\geq0
\]
where \(\delta\in(0, n + 1)\).

2. For \(n = 0, 1,\ldots\),
\[
U: (-1)^{n + 1}\frac{e^{-u}-\left(1-\frac{u}{1!}+\ldots+(-1)^{n}\frac{u^{n}}{n!}\right)}{u^{\delta}}=G_{n,\delta}(u)\quad\text{for}\quad u\geq0
\]
where \(\delta\in(n, n + 1)\).

The researchers develop a criterion by means of the Taylor formula (with the remainder in integral form) to determine whether a particular smooth function \(S(\delta)\) or \(U(\delta)\) is a lower or upper bound of \(s_{n}(\delta)\) or \(u_{n}(\delta)\). This enables the researchers to find relatively compact and simple upper and lower bounds of \(s_{n}\) and \(u_{n}\). Eventually, the researchers accurately estimate the positions of the maximum values of \(E_{n,\delta}(s)\) and \(G_{n,\delta}(u)\) as well as their maximum values \(ME_{n,\delta}\) and \(MG_{n,\delta}\).

In addition, the researchers show that \(ME_{n,\delta}\) and \(MG_{n,\delta}\) are convex as functions of \(\delta\), and determine their minimum values \(\hat{E}_{n}\) and \(\hat{G}_{n}\) as well as the positions of these minimum values \(\delta_{n,E}\) and \(\delta_{n,G}\).