File Exchange

image thumbnail

Control Functionals

version (16.2 KB) by Chris Oates
Implementation of the control functional method for computing Bayesian posterior integrals

1 Download

Updated 15 Mar 2017

View License

This is a compact implementation of the "control functional" method for Monte Carlo integration proposed in Oates, Chopin and Girolami, Journal of the Royal Statistical Society, Series B, 2017. The basic idea is that gradient information on the posterior distribution in Bayesian statistics can be used to better estimate integrals with respect to the posterior, via Stein's method.

Cite As

Chris Oates (2020). Control Functionals (, MATLAB Central File Exchange. Retrieved .

Comments and Ratings (3)

In 'control_func.m', isn't line 106 missing a multiplicative factor of n(or m in paper)? It doesn't make much of difference as lambda is set to be 10e-10, but just wondering after looking at the equation in the paper.


Fixed a typo in the pre-amble.

Now works in dimension d = 1,2,3 or 4.

Improved the example usage in the pre-amble so that it generalises to d = 2,3.

Fixed the numerical regulariser to match the method given in the paper. Thanks to Min Hyung Kang!

In version 1.0 the likelihood function was incorrect. The wrong covariance matrix was used. This is now fixed.

Renamed zip folder

Included description of the main function.

MATLAB Release Compatibility
Created with R2016a
Compatible with any release
Platform Compatibility
Windows macOS Linux