bilby.bilby_mcmc.sampler.BilbyPTMCMCSampler

class bilby.bilby_mcmc.sampler.BilbyPTMCMCSampler(convergence_inputs, pt_inputs, proposal_cycle, pt_rejection_sample, pool, use_ratio, evidence_method, initial_sample_method, initial_sample_dict, normalize_prior=True)[source]

Bases: object

__init__(convergence_inputs, pt_inputs, proposal_cycle, pt_rejection_sample, pool, use_ratio, evidence_method, initial_sample_method, initial_sample_dict, normalize_prior=True)[source]

__call__(*args, **kwargs): Call self as a function.

Methods

`__init__`(convergence_inputs, pt_inputs, ...)
`adapt_temperatures`()	Adapt the temperature of the chains
`compute_evidence`(outdir, label[, make_plots])
`compute_evidence_per_ensemble`(method, kwargs)
`ensemble_step`()
`get_initial_betas`()
`sampler_list_by_column`(column)
`set_convergence_inputs`(convergence_inputs)
`set_pt_inputs`(pt_inputs)
`setup_sampler_dictionary`(convergence_inputs, ...)
`step_all_chains`()
`stepping_stone_evidence`(ptchain, outdir, label)	Compute the evidence using the stepping stone approximation.
`swap_tempered_chains`()
`thermodynamic_integration_evidence`(ptchain, ...)	Computes the evidence using thermodynamic integration

Attributes

`evaluations`
`ln_z`
`ln_z_err`
`minimum_index`
`nsamples`
`nsamples_last`
`nsamples_nocache`
`position`
`primary_sampler`
`rejection_sampling_count`
`sampler_list`	A list of all individual samplers
`sampler_list_of_tempered_lists`
`samples`
`tau`
`tempered_sampler_list`
`zerotemp_sampler_list`

adapt_temperatures()[source]

Adapt the temperature of the chains

Using the dynamic temperature selection described in arXiv:1501.05823, adapt the chains to target a constant swap ratio. This method is based on github.com/willvousden/ptemcee/tree/master/ptemcee

property sampler_list: A list of all individual samplers

stepping_stone_evidence(ptchain, outdir, label, make_plots=True)[source]

Compute the evidence using the stepping stone approximation.

See https://arxiv.org/abs/1810.04488 and https://pubmed.ncbi.nlm.nih.gov/21187451/ for details.

The uncertainty calculation is hopefully combining the evidence in each of the steps.

Returns:

ln_z: float: Estimate of the natural log evidence
ln_z_err: float: Estimate of the uncertainty in the evidence

thermodynamic_integration_evidence(ptchain, outdir, label, make_plots=True)[source]

Computes the evidence using thermodynamic integration

We compute the evidence without the burnin samples, no thinning