search

dflemin3 / approxposterior

A Python package for approximate Bayesian inference and optimization using Gaussian processes
https://dflemin3.github.io/approxposterior/
MIT License
41 stars 9 forks source link

user-defined emcee samplers #12

Closed dflemin3 closed 5 years ago

dflemin3 commented 6 years ago

User should be able to supply their own initialized emcee (http://dfm.io/emcee/current/, the MCMC implementation approxposterior is currently set up to use) sampler object. The code is mostly there for a user to provide an initialized emcee sampler object, but it needs to be finalized. The user will need to pass None, or some dummy function, for the sampler's likelihood function since approxposterior will replace it with the GP loglikelihood function.

dflemin3 commented 5 years ago

Implemented for release 0.2. User can now pass dictionaries with all parameters for emcee.EnsembleSampler object and it's run_mcmc and sample methods.