Hello Christoph. Thanks for all your contributions to XAI.
One question on the "standard" conditional TreeSHAP algorithm applied to a tree-based model that is additive in covariate X (e.g., by using interaction constraints in a boosted trees model, or a boosted tree stump).
Is it correct to interpret the resulting SHAP dependence plot of a covariate X as being "Ceteris Paribus"?
mayer79
commented
2 years ago
Hello Christoph. Thanks for all your contributions to XAI.
One question on the "standard" conditional TreeSHAP algorithm applied to a tree-based model that is additive in covariate X (e.g., by using interaction constraints in a boosted trees model, or a boosted tree stump).
Is it correct to interpret the resulting SHAP dependence plot of a covariate X as being "Ceteris Paribus"?