Add default baseline parameter values for SHAP and Integrated Gradients

softwaremill / FoXAI

The open-source library for explainable AI. Generic and easy to integrate with PyTorch.

Apache License 2.0

30 stars 1 forks source link

Add default baseline parameter values for SHAP and Integrated Gradients #119

Open adamjankaczmarek opened 11 months ago

adamwawrzynski commented 11 months ago

I wonder if it is a good solution to leave such values as default. Baselines strongly influence the results of such an explanation, and I think it would be good for the user to be aware of what method he is using. Maybe enforcing this value explicitly is not the best solution, but a warning already might be. WDYT?