PROFESSIONAL-MACHINE-LEARNING-ENGINEER Question #246: Real Exam Question with Answer & Explanation

Question

You are using Vertex AI and TensorFlow to develop a custom image classification model. You need the model's decisions and the rationale to be understandable to your company's stakeholders. You also want to explore the results to identify any issues or potential biases. What should you do?

Options

• A1. Use TensorFlow to generate and visualize features and statistics.
• B1. Use TensorFlow Profiler to visualize the model execution.
• C1. Use Vertex Explainable AI to generate example-based explanations.
• D1. Use Vertex Explainable AI to generate feature attributions. Aggregate feature attributions over

Explanation

To make model decisions understandable and identify issues or biases in a custom image classification model, using Vertex Explainable AI to generate example-based explanations is effective.

Common mistakes.

• A. Generating and visualizing features and statistics with TensorFlow helps with data understanding, but doesn't directly explain why the model made a specific prediction or its rationale.
• B. TensorFlow Profiler helps in optimizing model training performance by visualizing execution, but it does not provide explanations for the model's predictions or help identify biases.
• D. While Vertex Explainable AI's feature attributions are useful for understanding feature importance, example-based explanations often provide more intuitive insights into why an image was classified a certain way for stakeholders and identifying specific decision issues.

Concept tested. Model explainability and bias detection with Vertex Explainable AI

Reference. https://cloud.google.com/vertex-ai/docs/explainable-ai/example-based-explanations

No community discussion yet for this question.