Wed, 24.03.2021 17:30

Introduction to Quasi-Monte Carlo Sampling - Part 2

Speaker: Art Owen (Stanford University)
Type: Group Seminar

Ask for zoom-link via officesfb(at)jku.at

Lecture 2 will cover randomized QMC (RQMC), primarily scrambling of digital nets. When it is important to have an accurate answer it is also important to verify somehow that the answer was accurate. Randomization provides a good approach to uncertainty quantification. Using the functional ANOVA it is possible to find properties of RQMC estimates (both asymptotic and finite sample) and to see that integrands dominated by low dimensional features are favorable to RQMC. RQMC can handle singular integrands and other functions that are not of bounded variation. Very recent work, joint with Daniel Rudolf, shows that scrambled nets consistently estimate integrals with 1 plus epsilon finite moments.