Model Interpretability and Explainability Questions

Discuss techniques for understanding what models learn: attention visualization, feature importance methods (SHAP, LIME), saliency maps, concept-based explanations. Understand the difference between post-hoc explainability and inherent interpretability. Discuss trade-offs with model complexity.

MediumTechnical

71 practiced

Explain Class Activation Mapping (CAM) and Grad-CAM for convolutional neural network interpretability. Describe architectural constraints required for CAM, how Grad-CAM generalizes CAM using gradients, steps to produce heatmaps, and limitations of these techniques when used for localization tasks.

EasyTechnical

87 practiced

List and briefly describe five feature importance techniques commonly used in practice (for example: mean decrease impurity, permutation importance, SHAP, LIME, coefficient magnitude). For each technique, note one advantage and one common pitfall when applied to real-world tabular data containing correlated features and missing values.

MediumTechnical

85 practiced

Write a Python function integrated_gradients(model, baseline, input_tensor, target_class, steps=50) using TensorFlow 2.x eager execution that returns a NumPy array of attributions with same shape as input_tensor. The model is a Keras model returning logits or probabilities. Explain how you handle multi-class outputs and baseline selection for images versus tabular data, and mention numerical stability considerations.

MediumBehavioral

74 practiced

Tell me about a time you explained a complex model's behavior to non-technical stakeholders or a business owner. Describe the audience, the specific model behavior you had to explain, the visualization or artifacts you used, how you handled pushback or misunderstandings, and what the outcome or decision was.

EasyTechnical

91 practiced

List and explain five practical limitations of LIME and SHAP when used in production environments. Consider stability/variance, handling correlated features, distribution shift, runtime cost, and the risk of producing misleading attributions. For each limitation suggest a practical mitigation or alternative.

Unlock Full Question Bank

Get access to hundreds of Model Interpretability and Explainability interview questions and detailed answers.

Join thousands of developers preparing for their dream job.