A metrical lower bound on the star discrepancy of digital sequences (Q2449427)

From MaRDI portal
scientific article
Language Label Description Also known as
English
A metrical lower bound on the star discrepancy of digital sequences
scientific article

    Statements

    A metrical lower bound on the star discrepancy of digital sequences (English)
    0 references
    0 references
    8 May 2014
    0 references
    Let \(\mathcal{S}=(\mathbf{x}_n)_{n\geq 0}\) be an infinite sequence in the \(s\)-dimensional unit cube \([0,1)^s\). For \(\mathbf{x}=(x_1,\dots, x_s)\in [0,1)^s\) the local discrepancy of \(\mathcal{S}\) is defined by \[ D(\mathbf{x}, N)=\#\{0\leq n\leq N: \mathbf{x}_n\in[\mathbf{0},\mathbf{x})=\prod_{j=1}^s[0,x_j)\}-Nx_1\cdots x_s. \] The star discrepancy is defined by \(D_N^*(\mathcal{S})=\|D(\mathbf{x},N)\|_{\infty}\). Let \(q\) be a prime number and let \(\mathbb{Z}_q: =\{0, 1, \dots, q-1\}\) equipped with arithmetic operations modulo \(q\). Let \(\mathcal{S}(C_1,\dots, C_s)\) be a digital sequence with generating matrices \(C_1,\dots , C_s\in\mathbb{Z}_q^{\mathbb{N}\times\mathbb{N}}\). A probability measure \(\mu_s\) on \((\mathbb{Z}_q^{\mathbb{N}\times\mathbb{N}})^s\) is introduced. In this paper the authors show that for \(\mu_s\)-almost all \(s\)-tuple \((C_1,\dots , C_s)\in (\mathbb{Z}_q^{\mathbb{N}\times\mathbb{N}})^s\) the digital sequence \(\mathcal{S}(C_1,\dots, C_s)\) has the star discrepancy satisfying \[ D_N^*(\mathcal{S}(C_1,\dots, C_s))\geq c(q,s)(\log N)^s\log\log N \] for infinitely many \(N\in \mathbb{N}\). This means that a better metrical upper bound of \textit{G. Larcher} [Lect. Notes Stat. 127, 109--123 (1997; Zbl 1109.65306)] is possible up to some \(\log\log N\) term.
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    discrepancy
    0 references
    metrical lower bound
    0 references
    digital sequence
    0 references
    uniform distribution
    0 references
    0 references
    0 references