hypothesis-kernel-analysis

Code for the "hypothesis kernel analysis" project.

Order to run scripts

Download the data from Figshare and store in data/. These data represent the preprocessed data.

The data-driven models (per split) are estimated by running:

4. src/analysis/estimate_datadriven_models.py

To convert the mappings into 2D embedding matrices that the kernel classifier can use, run:

5. src/analysis/convert_mappings_to_matrix.py

Finally, run the three-stage analysis:

6. src/analysis/prediction_analysis.py
7. src/analysis/explanation_analysis.py
8. src/analysis/exploration_analysis.py

To reproduce the paper's figures, execute the following notebook:

9. notebooks/figures.ipynb