This code archive contains the relevant software artefacts associated with the
Masters' project titled "A Statistical Measure of the Quality of Synthetic Data"

All code herein is implemented in Python3.

To run the DP-WGAN implementation, simply enter:

$ cd ./generative_models/dp_wgan

Where . represents the top-level directory of this archive. From there, run the
command:

$ python3 run.py

to execute the code. If you wish to change parameters, this must be done from
within run.py, or code associated with the generative mechanism.

To run the post-hoc classifier experiments, enter:

$ cd ./posthoc_classifiers

where . represents the top-level directory of this archive. From there, run the
command:

$ python3 main_test.py

to execute the code. If you wish to run the neural network test instead, run:

$ python3 nn_test.py

to execute the code. If you wish to alter parameters of the test, this must
likewise be done by editing either main_test.py, nn_test.py, or their associated
dependencies.

If you wish to plot the experiment data, run:

$ python3 plotting.py

If you wis to plot the data associated with the shuffling experiment, please run

$ python3 plotting_shuffles.py

instead.

The directory ./synth_results contains synthetic data generated by the the
following mechanisms: DP-Copula, DataSynthesizer, PATE-GAN, and DP-WGAN.
Each directory of results contains a file named method.txt which references
the method and provides a link to the associated github repository. The
exception to this is DP-WGAN, which by virtue of being implemented by the
student is included in this archive. The other three generation mechanisms 
have not been included here.

The directory ./posthoc_results contains .csv files corresponding to the
outputs of the experiments in the ./posthoc_classifiers, for ease of
verifying experimental results.