EmceeSampler

class ssapy.rvsampler.EmceeSampler(probfn, initializer, nWalker=50)[source][source]

Bases: object

A sampler built on the emcee package.

The emcee packages implements the Goodman and Weare (2010) affine-invariant sampler. This is often an efficient sampler to use when selecting a proposal distribution is not simple.

Parameters:

  • probfn (Callable) – Callable that accepts sampling parameters p and returns posterior probability.

  • initializer (Callable) – Callable that accepts number of desired initial samples and returns samples. Note, the initial samples should be (at least mostly) unique.

  • nWalker (int) – Number of ‘walkers’ to use in the Goodman & Weare algorithm. This should generally be at least 12 for the 6-dimensional problem.

probfn[source]
initializer[source]
nWalker[source]
sample(nBurn, nStep)[source][source]

Generate samples, first discarding nBurn steps, and then keeping nStep steps.

Methods Summary

sample([nBurn, nStep])

Generate samples.

Methods Documentation

sample(nBurn=1000, nStep=500)[source][source]

Generate samples.

Parameters:

  • nBurn (int) – Number of initial steps to take but discard.

  • nStep (int) – Number of subsequent steps to keep and return.

Returns:

  • chain (array (nStep, nWalker, 6)) – Generated samples. Columns are [x, y, z, vx, vy, vz].

  • lnprob (array (nStep, nWalker)) – The log posterior probabilities of the samples.

  • lnprior (array(nStep, nWalker)) – The log prior values for each step.