bayesian-fuelcycle
This repository contains code for using the Cyclus simulation software (https://fuelcycle.org) inside a likelihood function used by a Markov Chain Monte Carlo (MCMC) inference process driven by PyMC3.
This sounds complicated.
However, it is somewhat easy. For technical details, please refer to my bachelor's thesis at
thesis/thesis.pdf
, especially to its third and fourth sections.
The directories here are organized as follows:
-
cluster/
: scripts for running different inference scenarios on the RWTH SLURM cluster. -
cyclus/
: Disregard; only used for testing the Cyclus integration. -
docs/
: Presentations I've held about this topic. -
log/
: A log of all inference runs as well as my work log. -
notebooks/
: Supporting code in Jupyter notebooks. -
result_data/
: CDF files containing PyMC3 samples. -
sampling/
: The core code for running inferences using Cyclus. -
blackbox/
: Core code for blackbox likelihood function integration with PyMC3. -
cyclus_db/
: Extraction of results from Cyclus' SQLite simulation databases. -
models/
: Specific Cyclus models -
visualize/
: Evaluation of collected samples. -
singularity/
: Notes on running Cyclus inside a Singularity container. -
thesis/
: My thesis.
Not Ready for Reuse
Parts of this code are still under development by the authors, notably the code
and models in sampling/models/pakistan
, in preparation of a presentation and
publication in a journal or conference proceedings.