Integrated Gradients

After instantiating the package, run:

include("IntegratedGradients.jl")

and then:

IntegratedGradients.test_integrated_gradients()

to run a simple test that will return something similar to:

Counterfactual Training: 2.352, 29.25
abs(unprotected) / abs(protected) = 12.44

Conventional Training: 3.931, 4.768
abs('unprotected') / abs('protected') = 1.213