Question 21
Domain 3: Train and evaluate modelsMelinda is looking into architecture-appropriate SHAP algorithms and comes to you for clarification about which explainer uses an architecture-appropriate SHAP algorithm to interpret a model. Which of the following should you tell her is the best choice?
Correct answer: B
Explanation
TabularExplainer is designed for tabular data and uses architecture-appropriate SHAP methods to explain model predictions. SHAP explains outputs by assigning feature contributions, and this explainer is the standard choice when the model input is structured in columns rather than images or text.
Why each option is right or wrong
A. MimicExplainer
B. TabularExplainer
SHAP’s model-specific explainers are selected by data modality, and the tabular variant is the one built for structured, column-based inputs. In the SHAP library, `shap.explainers.Tabular` (formerly `shap.TabularExplainer`) is the explainer used for tabular models and applies the appropriate SHAP algorithm for that architecture, whereas image/text explainers are reserved for those data types.
C. RestExplainer
D. PFIExplainer