On integration methods based on scrambled nets of arbitrary size

Abstract: We consider the problem of evaluating

I (v a r p h i) := i n t_{[0, 1)^{s}} v a r p h i (x) d x

for a function

v a r p h i i n L^{2} [0, 1)^{s}

. In situations where

I (v a r p h i)

can be approximated by an estimate of the form

N^{- 1} s u m_{n = 0}^{N - 1} v a r p h i (x^{n})

, with

{x^{n}}_{n = 0}^{N - 1}

a point set in

[0, 1)^{s}

, it is now well known that the

O_{P} (N^{- 1 / 2})

Monte Carlo convergence rate can be improved by taking for

{x^{n}}_{n = 0}^{N - 1}

the first

N = l a m b d a b^{m}

points,

l a m b d a i n 1, d o t s, b - 1

, of a scrambled

(t, s)

-sequence in base

b g e q 2

. In this paper we derive a bound for the variance of scrambled net quadrature rules which is of order

o (N^{- 1})

without any restriction on

N

. As a corollary, this bound allows us to provide simple conditions to get, for any pattern of

N

, an integration error of size

o_{P} (N^{- 1 / 2})

for functions that depend on the quadrature size

N

. Notably, we establish that sequential quasi-Monte Carlo (M. Gerber and N. Chopin, 2015, emph{J. R. Statist. Soc. B, to appear.}) reaches the

o_{P} (N^{- 1 / 2})

convergence rate for any values of

N

. In a numerical study, we show that for scrambled net quadrature rules we can relax the constraint on

N

without any loss of efficiency when the integrand

v a r p h i

is a discontinuous function while, for sequential quasi-Monte Carlo, taking

N = l a m b d a b^{m}

may only provide moderate gains.

Recommendations

Cites work

Cited in

(11)

This page was built for publication: On integration methods based on scrambled nets of arbitrary size