UltraNest uses a hybrid RadFriends/MCMC method for exploring the parameter space. See the accompaning paper for details.

RadFriends is a method to sample from the neighbourhood of existing live points. This is highly efficient in the beginning of the integration, and for low-dimensional, peculiar parameter spaces. It also guarantees perfect uniform sampling. Once this method becomes inefficient, the algorithm switches to MCMC. The number of steps determines the efficiency (e.g. accept point after 100 steps = 1% efficiency). Still, RadFriends is used to narrow the MCMC proposal.

Call ultranest to run ultranest.

Download the code from the repository at https://github.com/JohannesBuchner/UltraNest/ (subfolder ultranest)

Author: Johannes Buchner

Date: 2014