Support SAGE values similar to SHAP

[jckkvs]

I am also interested in the SAGE value, which measures the contribution to global prediction accuracy, compared to SHAP, which measures the importance of local predictions.
We believe that if SRI decomposition can contribute to prediction accuracy, it will be an important advance in terms of model interpretability and feature selection.

On the other hand, I understand that it is not easy to support SAGE, which is a different concept, but is it likely to be realized?

https://pypi.org/project/sage-importance/
https://iancovert.com/blog/understanding-shap-sage/