Multidimensional quasi-Monte Carlo methods (Q1314409)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Multidimensional quasi-Monte Carlo methods |
scientific article |
Statements
Multidimensional quasi-Monte Carlo methods (English)
0 references
3 March 1994
0 references
Four types of \(s\)-dimensional sequences of the so-called quasi-random numbers are quoted, namely the sequences of \textit{J. Halton} [Numer. Math. 2, 84-90 (1960; Zbl 0090.345)], \textit{I. M. Sobol'} [USSR Comput.Math. Math. Phys. 7, No. 4, 86-112 (1967), translation from Zh. Vychisl. Mat. Mat. Fiz. 7, 784-802 (1967; Zbl 0185.411)], \textit{H. Faure} [Acta Arith. 41, 337-351 (1982; Zbl 0488.10054)] and \textit{H. Niederreiter} [Monatsh. Math. 104, 273-337 (1987; Zbl 0626.10045)]. It is well known that all these sequences are equidistributed in \([0,1]s\) and are utilized mostly for the calculation of multiple integrals. The constructions of Halton, Faure and Niederreiter sequences are recalled and their more important properties are illustrated. It is observed that the Niederreiter sequences can be regarded as generalization of those of Sobol' and Faure. In particular the discrepancy \(D(N)\) of the subsequence of the first \(N\) elements is considered for every type of sequences. As is known, for these discrepancies inequalities of the following type do hold: \(D(N) \leq K_ s(\log N)^ s + O((\log N)^{s - 1})\), where \(K_ s\) depends on the sequence considered. In the paper the expressions of the \(K_ s\) are reported and compared. The author hints at the possibility of obtaining new sequences by a mixage of sequences of the above mentioned types and outlines a project in this direction. The resulting sequence should be ``telescopic'', i.e. it would be possible, from a sequence of dimension \(s\), to obtain another sequence of dimension \(s + 1\) by joining only the new coordinate at every point.
0 references
multidimensional quasi-Monte Carlo methods
0 references
random number generation
0 references
quasi-random numbers
0 references
discrepancies inequalities
0 references
0 references