Nested Sampling: Difference between revisions
No edit summary |
No edit summary |
||
| Line 1: | Line 1: | ||
Nested sampling is an alternative method of MCMC sampling used to make inferences about models from data. | Nested sampling is an alternative method of MCMC sampling used to make inferences about models from data. | ||
Importantly, nested sampling approaches inference with the goal of calculating the Bayesian Evidence for model comparison. However, the joint posterior for the model parameters are also generated in the process. | |||
A commonly used implementation of nested sampling is MultiNest, which in addition to the vanilla algorithm also allows multiple likelihood modes to be identified during the process of inference. This may be useful for example in source finding or spectral line modelling. | |||
The PyMultiNest wrapper can be found at phttps://johannesbuchner.github.io/PyMultiNest/]. Note that MultiNest is notoriously challenging to compile and install, and so depending on how much time you have you may wish to just use the installed version on Glamdring. | |||
Slides from a graduate lecture given by James Allison can be found here [[File:Nested sampling lecture 2016.pdf ]] | Slides from a graduate lecture given by James Allison can be found here [[File:Nested sampling lecture 2016.pdf ]] | ||
Useful references: | |||
[https://www.inference.org.uk/bayesys/] | |||
[https://ui.adsabs.harvard.edu/abs/2008MNRAS.384..449F/] | |||
[https://ui.adsabs.harvard.edu/abs/2009MNRAS.398.1601F/] | |||
[https://ui.adsabs.harvard.edu/abs/2019OJAp....2E..10F] | |||
Revision as of 17:02, 6 November 2020
Nested sampling is an alternative method of MCMC sampling used to make inferences about models from data.
Importantly, nested sampling approaches inference with the goal of calculating the Bayesian Evidence for model comparison. However, the joint posterior for the model parameters are also generated in the process.
A commonly used implementation of nested sampling is MultiNest, which in addition to the vanilla algorithm also allows multiple likelihood modes to be identified during the process of inference. This may be useful for example in source finding or spectral line modelling.
The PyMultiNest wrapper can be found at phttps://johannesbuchner.github.io/PyMultiNest/]. Note that MultiNest is notoriously challenging to compile and install, and so depending on how much time you have you may wish to just use the installed version on Glamdring.
Slides from a graduate lecture given by James Allison can be found here File:Nested sampling lecture 2016.pdf