Uncertainty Quantification Working Group
January 10, 11:30 AM, CNLS Conf. Room, TA-3, Bldg. 1690

Effective Error Estimates for Quasi-Monte-Carlo Computations

Tony Warnock, CCS-3 (ttw@lanl.gov )


Quasi-Monte-Carlo methods are based on sampling over a low-discrepancy point set rather than a randomly chosen point set as is used in traditional Monte Carlo. For a comparison of the two sampling methods, see below. Although Quasi-Monte-Carlo methods have a superior convergence rate than traditional Monte-Carlo methods, they have suffered from a lack of an effective error estimate. An effective error estimate is described based on independent replications of Quasi-Monte-Carlo computations.

Quasi-Monte-Carlo sampling