Welcome to LLNL’s IBIS Documentation!#

Interactive Bayesian Inference and Sensitivity, or IBIS, is a Python-based scientific tool for analyzing concurrent UQ simulations on high-performance computers. Using a simple, non-intrusive interface to simulation models, it provides the following capabilities:

  • generating parameter studies

  • generating one-at-a-time parameter variation studies

  • sampling high dimensional uncertainty spaces

  • constructing surrogate models

  • performing sensitivity studies

  • performing statistical inferences

  • estimating parameter values and probability distributions

IBIS#

IBIS has been used for simulations in the domains of Inertial Confinement Fusion, National Ignition Facility experiments, climate, as well as other programs and projects. The ibis package is designed to be used after a number of simulations have run to completion. This package is used to predict the results of future simulation runs and to assess the sensitivity each input parameter has on the output.

The ibis package works with Python 3 and is made up of 7 modules:

  • filter

  • likelihoods

  • mcmc

  • mcmc_diagnostics

  • sensitivity

  • pce_model

  • plots

Installation:

pip install llnl-ibis

Demo usage :

from ibis import mcmc, pce_model

Indices and tables#