We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Mathematics > Probability

Title: Monte Carlo with Determinantal Point Processes

Abstract: We show that repulsive random variables can yield Monte Carlo methods with faster convergence rates than the typical $N^{-1/2}$, where $N$ is the number of integrand evaluations. More precisely, we propose stochastic numerical quadratures involving determinantal point processes associated with multivariate orthogonal polynomials, and we obtain root mean square errors that decrease as $N^{-(1+1/d)/2}$, where $d$ is the dimension of the ambient space. First, we prove a central limit theorem (CLT) for the linear statistics of a class of determinantal point processes, when the reference measure is a product measure supported on a hypercube, which satisfies the Nevai-class regularity condition, a result which may be of independent interest. Next, we introduce a Monte Carlo method based on these determinantal point processes, and prove a CLT with explicit limiting variance for the quadrature error, when the reference measure satisfies a stronger regularity condition. As a corollary, by taking a specific reference measure and using a construction similar to importance sampling, we obtain a general Monte Carlo method, which applies to any measure with continuously derivable density. Loosely speaking, our method can be interpreted as a stochastic counterpart to Gaussian quadrature, which, at the price of some convergence rate, is easily generalizable to any dimension and has a more explicit error term.
Comments: 58 pages, 6 figures. To appear in Annals of Applied Probability
Subjects: Probability (math.PR); Classical Analysis and ODEs (math.CA); Computation (stat.CO); Methodology (stat.ME)
Cite as: arXiv:1605.00361 [math.PR]
  (or arXiv:1605.00361v2 [math.PR] for this version)

Submission history

From: Adrien Hardy [view email]
[v1] Mon, 2 May 2016 06:24:06 GMT (3510kb,D)
[v2] Sat, 15 Jun 2019 11:20:01 GMT (4182kb,D)

Link back to: arXiv, form interface, contact.