26th April 2004 — Nested Sampling

David Mackay will present Nested Sampling (nee Box sampling), John Skilling's new Monte Carlo method for evaluating the normalising constant of a distribution. This is obviously important as the "evidence" in a Bayesian problem. The method can optionally make predictions and/or draw samples.

Get the paper (and optionally sample code and a lower level description) from the right-hand side of http://www.inference.phy.cam.ac.uk/bayesys/

If you decide to look through the code, bear in mind that it corresponds to the earlier paper included in the archive: "box sampling". This is nested sampling, but some implementation details are different.