This folder contains all notebooks used from downloading the data to generating the results. We list them below in the recommended order, along with a short description
- download_data.ipynb Is for automatically downloading
the data needed for using the rest of the notebooks. Please read the relevant
DATA.md
file in the project root for details on downloadable data. - preprocess_data.ipynb Is for preprocessing the data for training and other processes.
- train_models.ipynb Is for training the models.
- pairing.ipynb Is for generating the
seed_set_pairings.csv
available in the project root. As can be seen, these are manual pairings and do not necessarily reflect like sensible or good pairings in terms of bias subspaces - fig_1.ipynb Is for producing Figure 1 in our report, which is a replication and extension of Fig. 4 in the original paper. The same notebook is used to produce Figure A.2 of our report, which is a further extension.
- fig_2.ipynb Is for producing Figure 2 in our report, which is a replication of Fig. 3 in the original paper.
- fig_3.ipynb Is for producing Figure 3 in our report, which is a replication of Fig. 2 in the original paper.
- fig_4.ipynb Is for producing Figure 4 in our report, which is a replication and extension of Fig. 5 in the original paper.
- fig_a.1.ipynb Is for producing Figure A.1 of our report which is a replication of Figure 1 in the original paper.
- tab_a.1.ipynb Is for producing Table A.1 in our report, which is a replication of Table 2 in the original paper
- tab_a.2.ipynb Is for producing Table A.2 in our report, which is a replication of Table 2 in the original paper