rxmc

rxmc is an orchestration layer for Bayesian calibration of reaction models to large data sets with flexible, composable covariance modeling.

It is built around two complementary workflows:

1. External-sampler orchestration via CalibrationConfig for drivers such as black-box-bayes.

2. In-package end-to-end prototyping via Walker for smaller problems where you want to run the full MCMC workflow locally.

The package composes curated experimental data (Observation), model predictions (PhysicalModel), uncertainty declared as additive covariance Term s (statistical, systematic, unknown-noise, and Gaussian-process discrepancy modes), maximal blocks of mutually-correlated data (Constraint), and full calibration problems (Evidence).

Contents

• Installation
• Design: the stacked covariance model
• A ground-up rxmc: declare, then compile
• Recipes
• Bugs and inconsistencies found during the architecture review
• API Reference
• Examples