Improvements on the star discrepancy of (t,s)-sequences
نویسندگان
چکیده
The term low-discrepancy sequences is widely used to refer to s-dimensional sequences X satisfying the bound D∗(N,X) ≤ cs(logN) + O((logN)s−1), where D∗ denotes the usual star discrepancy. In this article, we are concerned with (t, s)-sequences in base b, one of the most famous families of low-discrepancy sequences along with Halton sequences. The constants cs for (t, s)-sequences were first computed by Sobol’ and Faure in special cases and then achieved in general form by Niederreiter in the eighties. Then, quite recently, Kritzer improved these constants for s ≥ 2 by a factor 1 2 for an odd base and b 2(b+1) for an even base b ≥ 4. Our aim is to further improve the result in the case of an even base by a factor 2(b+1) b ( b−1 b )s−1 (s ≥ 2), hence obtaining the ratio 3 2s−1 for b = 2. Combining this estimate with best known t-values from the database MinT, we obtain new values for Niederreiter-Xing sequences where base 2 recovers supremacy on base 3 (in Kritzer’s paper). Our proof relies on a method of Atanassov to bound the discrepancy of Halton sequences. We also investigate (t, 1)-sequences for which the approach of Kritzer does not work.
منابع مشابه
New star discrepancy bounds for ( t , m , s ) - nets and ( t , s ) - sequences
In this paper, we derive new general upper bounds on the star discrepancy of (t,m, s)-nets and (t, s)-sequences. These kinds of point sets are among the most widely used in quasi-Monte Carlo methods for numerical integration. By our new results, we improve on previous discrepancy bounds and prove a conjecture stated by the second author in an earlier paper.
متن کاملDiscrepancy bounds for low-dimensional point sets
The class of (t,m, s)-nets and (t, s)-sequences, introduced in their most general form by Niederreiter, are important examples of point sets and sequences that are commonly used in quasi-Monte Carlo algorithms for integration and approximation. Low-dimensional versions of (t,m, s)-nets and (t, s)-sequences, such as Hammersley point sets and van der Corput sequences, form important sub-classes, ...
متن کاملImprovements on Low Discrepancy One-dimensional Sequences and Two-dimensional Point Sets
We obtain significant improvements for the star discrepancy D∗ of generalized van der Corput sequences by means of linear digit scramblings (see Section 5.2 for the definition). We also find a new lower bound for the extreme discrepancyD of these sequences which permits to show that linearly-scrambled sequences are set in a good place among generalized van der Corput sequences. Finally, we deri...
متن کاملGeneralized Hofer-Niederreiter sequences and their discrepancy from an (u, e, s)-point of view
Recently Tezuka introduced the concept of (u, e, s)-sequences which generalizes (t, s)-sequences by Niederreiter. This generalization can be used to point out a deeper regularity of certain digital sequences, which can be exploited to obtain improvements on their discrepancy bounds. Earlier Larcher and Niederreiter introduced so-called (T , s)-sequences with similar consequences. In this paper ...
متن کاملDiscrepancy Bounds for Mixed Sequences
A common measure for the uniformity of point distributions is the star discrepancy. Let λ s denote the s-dimensional Lebesgue measure. Then the star discrepancy of a mul-tiset P = {p 1 ,. .. , p N } ⊂ [0, 1] s is given by D * N (P) := sup α∈[0,1] s λ s ([0, α)) − 1 N N k=1 1 [0,α) (p k) ; here [0, α) denotes the s-dimensional axis-parallel box [0, α 1) × · · · × [0, α s) and 1 [0,α) its charact...
متن کامل