dyPolyChord: dynamic nested sampling with PolyChord
Edward Higson · The Journal of Open Source Software · 2018
Nested sampling (Skilling, 2006) is a popular numerical method for calculating Bayesian evidences and generating posterior samples given some likelihood and prior.The initial development of the algorithm was targeted at evidence calculation, but implementations such as MultiNest (Feroz & Hobson, 2008;Feroz, Hobson, & Bridges, 2008;Feroz, Hobson, Cameron, & Pettitt, 2013) and PolyChord (W.J. Handley, Hobson, & Lasenby, 2015a, 2015b) are now used extensively for parameter estimation in scientific research (and in particular in astrophysics); see for example (Chua et al., 2018; DES Collaboration, 2018).Nested sampling performs well compared to Markov chain Monte Carlo (MCMC)-based alternatives at exploring multimodal and degenerate distributions, and the PolyChord software is well-suited to high-dimensional problems.Dynamic nested sampling (Higson, Handley, Hobson, & Lasenby, 2017) is a generalisation of the nested sampling algorithm which dynamically allocates samples to the regions of the posterior where they will have the greatest effect on calculation accuracy.This allows order-of-magnitude increases in computational efficiency, with the largest gains for high dimensional parameter estimation problems.