Low-Level API — Overview

BayesBay offers a low-level API for a highly customizable experience. Instead of creating one or more instances of ParameterSpace or Discretization to define a Parameterization as in the standard API (see Standard API — Overview), our low-level API allows users to write their own perturbation functions and define a Markov chain state via an arbitrary Python object. More specifically, the steps involved in a typical Bayesian inference defined through our low-level API involve the following:

1. Define an arbitrary log-likelihood function

2. Define arbitrary perturbation functions to propose a Markov chain state from the current one

3. Generate your initial Markov chain states

4. Define an instance of bayesbay.BaseBayesianInversion using the above objects

5. Run the inversion (bayesbay.BaseBayesianInversion.run())