This repository contains a copy of the code for running the experiments in the Scalable Extreme Deconvolution paper. At some point in the future we hope to tidy it up and refactor it into a proper package.
An export of the conda environment used for the experiments is provided in environment.yml.
For the baseline experiments, the original extreme deconvolution implementation will need to be installed manually and added to your PYTHONPATH.
Either install the deconv package in editable mode using pip install -e ., or otherwise add it to your PYTHONPATH.
The deconv/experiments/gaia/preprocessing.py script can convert a VO table file from a query on the DR2 source table using the Gaia Archive into a Numpy .npz file for use with the experiment scripts.
The scripts in deconv/experiments/gaia/ show how to fit the models to the .npz datafiles.
The bash scripts are specific to our cluster, but show the options used in the experiments.
deconv/experiments/checks/ contains some small toy experiments.