A combination of Monte Carlo and classical methods for evaluating multiple integrals

Haber, Seymour

doi:10.1090/s0002-9904-1968-11993-7

Cited by 8 publications

(1 citation statement)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recalling the discussion in Section 3.1, MCLSA integrates f exactly if it is a polynomial of degree at most k. Thus as long as V has full column rank, we obtain p = f , and hence the integral is exact. Such (randomized) integration formulae that provide exact results for polynomials of bounded degree were investigated by Haber [21,22], and are called stochastic quadrature formulae.…”

Section: Exact Degreementioning

confidence: 99%

Approximate and integrate: Variance reduction in Monte Carlo integration via function approximation

Nakatsukasa¹

2018

Preprint

View full text Add to dashboard Cite

Classical algorithms in numerical analysis for numerical integration (quadrature/cubature) follow the principle of approximate and integrate: the integrand is approximated by a simple function (e.g. a polynomial), which is then integrated exactly. In highdimensional integration, such methods quickly become infeasible due to the curse of dimensionality. A common alternative is the Monte Carlo method (MC), which simply takes the average of random samples, improving the estimate as more and more samples are taken. The main issue with MC is its slow (though dimension-independent) convergence, and various techniques have been proposed to reduce the variance. In this work we suggest a numerical analyst's interpretation of MC: it approximates the integrand with a constant function, and integrates that constant exactly. This observation leads naturally to MC-like methods where the approximant is a non-constant function, for example low-degree polynomials, sparse grids or low-rank functions. We show that these methods have the same O(1/ √ N ) asymptotic convergence as in MC, but with reduced variance, equal to the quality of the underlying function approximation. We also discuss methods that improve the approximation quality as more samples are taken, and thus can converge faster than O(1/ √ N ). The main message is that techniques in high-dimensional approximation theory can be combined with Monte Carlo integration to accelerate convergence.

show abstract

Section: Exact Degreementioning

confidence: 99%