pocomc-bilby

pocomc plugin for bilby.

This package provides a plugin for using pocomc with bilby via the sampler plugins interface. pocomc is Sequential Monte Carlo sampler that implements a specific flavour of SMC called Persistent Sampling. For more details, see the pocomc documentation.

Installation

Note: since pocomc depends on torch we recommend installing it first using the instructions on the PyTorch website.

pocomc-bilby can be installed via pip:

pip install pocomc-bilby

Note: this plugin only supports pocomc>=1.2.6.

Usage

Once installed, pocomc can be used as you would any other sampler in bilby:

import bilby

# Define likelihood & priors as normal
priors = ...
likelihood = ...

bilby.run_sampler(
    sampler="pocomc",
    ...
)

pocomc has two different types of settings, those specified when initializing the sampler and those specified when calling the run method. When using pocomc via bilby both types of settings can be passed to run_sampler and these will automatically be passed to the correct method.

For details about the various settings, see the pocomc documentation.

This plugin is also compatible with bilby_pipe, including support for checkpointing.

Attribution & Citation

If you use pocomc-bilby in your own work please cite the DOI for this package, corresponding paper and the citations for pocomc, see Attribution & Citation in the pocomc documentation