Google Professional Machine Learning Engineer
Get started today
Ultimate access to all questions.
Comments
Loading comments...
Ultimate access to all questions.
As a professional working for a textile manufacturer, you've developed a machine learning model aimed at detecting and classifying fabric defects from high-resolution images captured at the production line's end. The model has been trained to achieve high recall, ensuring that most defects are identified. However, to build trust among quality control inspectors, it's crucial to not only detect defects but also to explain the classifier's decision-making process clearly. The production environment requires that the explanation method must be computationally efficient to handle the high volume of images processed daily and must provide interpretable results to non-technical inspectors. Which method should you employ to elucidate the rationale behind your classifier's predictions while meeting these constraints? (Choose two correct options if E is available)
A
Employ K-fold cross-validation to assess the model's performance across various test datasets, ensuring robustness.
B
Leverage the Integrated Gradients method to calculate feature attributions for each image prediction efficiently, highlighting the areas of the image most influential to the prediction.
C
Apply PCA (Principal Component Analysis) to condense the original feature set into a more manageable and comprehensible subset, though it may obscure the interpretability of specific features.
D
Utilize k-means clustering to categorize similar images and evaluate cluster separation with the Davies-Bouldin index, focusing on grouping rather than explanation.
E
Implement both SHAP (SHapley Additive exPlanations) for global interpretability and LIME (Local Interpretable Model-agnostic Explanations) for local interpretability, ensuring comprehensive understanding across different scales.