English

Control Functionals for Quasi-Monte Carlo Integration

Computation 2016-04-04 v7

Abstract

Quasi-Monte Carlo (QMC) methods are being adopted in statistical applications due to the increasingly challenging nature of numerical integrals that are now routinely encountered. For integrands with dd-dimensions and derivatives of order α\alpha, an optimal QMC rule converges at a best-possible rate O(Nα/d)O(N^{-\alpha/d}). However, in applications the value of α\alpha can be unknown and/or a rate-optimal QMC rule can be unavailable. Standard practice is to employ αL\alpha_L-optimal QMC where the lower bound αLα\alpha_L \leq \alpha is known, but in general this does not exploit the full power of QMC. One solution is to trade-off numerical integration with functional approximation. This strategy is explored herein and shown to be well-suited to modern statistical computation. A challenging application to robotic arm data demonstrates a substantial variance reduction in predictions for mechanical torques.

Keywords

Cite

@article{arxiv.1501.03379,
  title  = {Control Functionals for Quasi-Monte Carlo Integration},
  author = {Chris. J. Oates and Mark Girolami},
  journal= {arXiv preprint arXiv:1501.03379},
  year   = {2016}
}

Comments

To appear at AISTATS 2016 (oral presentation)

R2 v1 2026-06-22T08:01:15.235Z