Average equidistribution and statistical independence properties of digital inversive pseudorandom numbers over parts of the period (Q2781225)
From MaRDI portal
| This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use this page instead for the normal view: Average equidistribution and statistical independence properties of digital inversive pseudorandom numbers over parts of the period |
scientific article; zbMATH DE number 1720974
| Language | Label | Description | Also known as |
|---|---|---|---|
| English | Average equidistribution and statistical independence properties of digital inversive pseudorandom numbers over parts of the period |
scientific article; zbMATH DE number 1720974 |
Statements
19 March 2002
0 references
uniform pseudorandom numbers
0 references
digital inversive method
0 references
average equidistribution behaviour
0 references
average statistical independence properties
0 references
star discrepancy
0 references
exponential sums
0 references
0 references
0 references
0 references
0.8605869
0 references
0.83542794
0 references
0.8255028
0 references
0.8172978
0 references
0.79510486
0 references
0.79422206
0 references
0.79266894
0 references
0.78804654
0 references
Average equidistribution and statistical independence properties of digital inversive pseudorandom numbers over parts of the period (English)
0 references
The author of this interesting paper concentrates on digital inversive pseudorandom numbers generators introduced by \textit{J. Eichenauer-Herrmann} and \textit{H. Niederreiter} [ACM Trans. Model. Comput. Simul. 4, 339-349 (1994; Zbl 0847.11038)]. After summarizing some known results and auxiliary results, he shows in Theorem~1 that in the digital inversive method the star discrepancy, on the average over the parameter \(\kappa\), has an order of magnitude at most \(N^{-1/2}(\log q)^s\) for any parameters \(\alpha\) and \(\beta\) defining the generator, provided the condition for the maximum possible period length is met. Moreover, it is shown that for any fixed parameters \(\alpha\) and \(\beta\), only a small percentage of the values of the parameter \(\kappa\) may lead to a star discrepancy with an order of magnitude greater than \(N^{-1/2}(\log q)^s\). On the other hand, Theorem~2 shows that, for any \(\alpha\) and \(\beta\), there exists values of the parameter \(\kappa\) such that the star discrepancy is of an order of magnitude at least \(N^{1/2}\), if \(N\) is not too close to \(q\).
0 references