This code allows to recreate the results from the paper "Explainability as statistical inference". 
Since there are a lot of different parameters to iterate over, we have multiple sets of launcher to consider
to recreate all the experiments.

ARTIFICIAL DATASETS :

To recreate figures from the Artificial dataset analysis, one should launch all 
the test_*.py in Examples/ArtificialRegression/experiment_artificial.

For instance :

python test_realx_subset_cste.py

will train LEX models with a surrogate constant imputation parameterizing with a subset sampling distribution 
by iterating over all the sampling rate, constant of imputation and the three datasets S1, S2, S3.

PANEL MNIST :

Similary, to recreate the results from section 4.2 on Panel MNIST, one should launch all 
the test_*.py in Examples/ImageExplainer/experiment_fashionandmnist.

To recreate the extended results from the appendix, one should launch all
the test_*.py in Examples/ImageExplainer/experiment_fashionandmnist_extended.

To recreate results with the original parameterization of the previously existing model used for the rebuttal, 
one should launch all the test_*.py in Examples/ImageExplainer/experiment_fashionandmnist_realparam

CELEBA SMILE :

To recreate the results from section 4.3 on CELEBA SMILE, one should launch all the
test_*.py in Examples/ImageExplainer/experiment_celeba_dataset


All the results will be store in csv files to analyse.
