This repo contains all the code implemented and described in chapter 4 of my MSc thesis. In order to run the code one needs all the standard python3 libraries and the NIFTy8.
All the models are implemented in the bipartite_model.py and confounder_model.py with their shared options implemented inside causal_model.py, which also contains the method for estimating model evidences (the _get_evidence(**args)). For changing the hyperparameters of the models themselves take a look into the config.json file. The data_processing_utilities.py contains the relevant setup information for configuring the code arguments (look into Parser). Main pyscript is do_bci_inference.py which does the actual tests.
An example of how to run the tests and select models / testcases is presented inside the evaluation_scripts/run.sh bash script. First copy it to the $ROOT folder containing the do_bci_inference.py script and run it. For any problems contact me at [email protected].