Research code for the paper "Sequential Design of Computer Experiments for the Computation of Bayesian Model Evidence"
A more organized implementation of the paper's method is available in my other repository bali. I strongly recommend using that implementation, unless you need to specifically reproduce the experiments from the article.
To run the code, you need the python packages
- numpy (1.16.4)
- scipy (1.3.0)
- emcee. (2.2.1)
The two scripts 01_computations.py and 02_computations.py run the experiments and generate output data. The script 03_plots.py processes the output data into tables that can be uses for creating the plots (with pgfplots).
- bbi contains the package for running sequential designs
- data contains the dataset for the first experiment and the dataset and input-output data for the second experiment
- ouput contains the output generated by the computations in 01_computations.py and 02_computations.py (content can be reproduced by running these)
- plotdata contains the files that are generated by 03_plots.py
Due to filesize the input-output-data of experiment 2 is not included in this repository. Please contact the authors, if you need the data.