Python Bayesian Inference Toolbox and Uncertainty Propagation

1. Bayesian calibration for parameter inference

   1. Experimental data or synthetic data, with one or several physical models.
   2. Markov chain Monte Carlo (MCMC) method using Metropolis-Hastings algorithms, Ito Stochastic differential equation.
   3. Compute correlation among input variables.
   4. Posterior predictive checks.

2. Uncertainty propagation

   1. Monte carlo or polynomial chaos methods.
   2. Using labelled distributions or propagate MCMC samples with distributions that can be correlated.

3. Sensitivity analysis

   1. From MCMC chains.
   2. Monte Carlo method or Kernel method.

1. Sensitivity analysis module still needs to be developped.
2. Couple PCE surrogate model to perform sensitivity analysis.

The code can be installed as a python package using the command:

python -m pip install git+https://github.com/jcoheur/pybitup    

Add @branch_name to install a particular branch from the git.

See examples in the pybitup-example repository.