Topic: explainability
-
SHAP attributions: reference values, output units and causal limits
Explain a particular prediction relative to a declared reference, track the output scale, and avoid interpreting predictive contributions as intervention effects.
Machine-readable: JSON