English

On Integration Methods Based on Scrambled Nets of Arbitrary Size

Computation 2015-06-09 v3 Numerical Analysis

Abstract

We consider the problem of evaluating I(φ):=[0,1)sφ(x)dxI(\varphi):=\int_{[0,1)^s}\varphi(x) dx for a function φL2[0,1)s\varphi \in L^2[0,1)^{s}. In situations where I(φ)I(\varphi) can be approximated by an estimate of the form N1n=0N1φ(xn)N^{-1}\sum_{n=0}^{N-1}\varphi(x^n), with {xn}n=0N1\{x^n\}_{n=0}^{N-1} a point set in [0,1)s[0,1)^s, it is now well known that the OP(N1/2)O_P(N^{-1/2}) Monte Carlo convergence rate can be improved by taking for {xn}n=0N1\{x^n\}_{n=0}^{N-1} the first N=λbmN=\lambda b^m points, λ{1,,b1}\lambda\in\{1,\dots,b-1\}, of a scrambled (t,s)(t,s)-sequence in base b2b\geq 2. In this paper we derive a bound for the variance of scrambled net quadrature rules which is of order o(N1)o(N^{-1}) without any restriction on NN. As a corollary, this bound allows us to provide simple conditions to get, for any pattern of NN, an integration error of size oP(N1/2)o_P(N^{-1/2}) for functions that depend on the quadrature size NN. Notably, we establish that sequential quasi-Monte Carlo (M. Gerber and N. Chopin, 2015, \emph{J. R. Statist. Soc. B, to appear.}) reaches the oP(N1/2)o_P(N^{-1/2}) convergence rate for any values of NN. In a numerical study, we show that for scrambled net quadrature rules we can relax the constraint on NN without any loss of efficiency when the integrand φ\varphi is a discontinuous function while, for sequential quasi-Monte Carlo, taking N=λbmN=\lambda b^m may only provide moderate gains.

Keywords

Cite

@article{arxiv.1408.2773,
  title  = {On Integration Methods Based on Scrambled Nets of Arbitrary Size},
  author = {Mathieu Gerber},
  journal= {arXiv preprint arXiv:1408.2773},
  year   = {2015}
}

Comments

27 pages, 2 figures (final version, to appear in The Journal of Complexity)

R2 v1 2026-06-22T05:26:47.575Z