Objective 3.5 Address Business Risks, Ethical Concerns, and Related Concepts in Training and Tuning AIP-210 Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

   Which of the following is a technique used to reduce bias in a machine learning model during training?

   • AUsing a larger batch size to stabilize gradient updates.
   • BReweighting training samples to give more importance to underrepresented groups.
   • CAdding more features to the model to capture more information.
   • DIncreasing the learning rate to speed up convergence.
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   Which of the following is a common business risk that can arise during the training and tuning of an ML model?

   • AThe training data is split into training, validation, and test sets to evaluate performance.
   • BThe model inadvertently uses a protected attribute (e.g., race or gender) as a proxy, leading to biased decisions.
   • CThe model's hyperparameters are tuned to maximize accuracy on the validation set.
   • DThe model's loss function is minimized using gradient descent during training.
   Reveal Solution Discussion Select an answer first
4. 8foundation · easy

   What is model drift in the context of ML monitoring?

   • AThe degradation of model performance over time due to changes in the underlying data distribution.
   • BThe intentional modification of model weights to improve accuracy.
   • CThe process of moving a model from a development environment to production.
   • DThe removal of outdated features from the model.
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A city uses an AI system to prioritize fire inspections. The system was trained on historical inspection data. Community advocates are concerned that the system may over-target low-income neighborhoods. The city wants to ensure that the system does not unfairly burden any community. Which action should they take?

   • AEvaluate the model using equalized odds to ensure that false positive rates are similar across neighborhoods.
   • BUse demographic parity to ensure that each neighborhood receives the same number of inspections.
   • CIncrease the model's threshold for inspections to reduce the number of inspections overall.
   • DRemove the neighborhood feature from the model to prevent bias.
   Reveal Solution Discussion Select an answer first
6. 10foundation · easy

   Which of the following is an example of monitoring for unintended societal impacts of a model?

   • AReviewing the model's feature importance values.
   • BTracking whether the model's error rates differ across demographic groups over time.
   • CMeasuring the model's inference latency on a daily basis.
   • DChecking the model's accuracy on a holdout test set.
   Reveal Solution Discussion Select an answer first