Oversee AI/ML Model QA/QC (e.g., Configuration Management, Model Performance) PMI-CPMAI Practice Questions (Page 4)

Questions 16–20

1. 16application · medium

   A machine learning engineer is about to start a new experiment to tune hyperparameters for a random forest model. The team uses Git for code and DVC for data. What should the engineer do to ensure the experiment is reproducible?

   • AOnly commit the code and rely on the current data version in the shared drive
   • BCreate a Git tag for the code and a DVC tag for the data, and log the hyperparameters in a separate file
   • CUse a spreadsheet to record the hyperparameters and data file names
   • DCommit the code, data, and hyperparameters to the same Git repository
   Reveal Solution Discussion Select an answer first
2. 17application · medium

   A QA lead is establishing QA procedures for a new AI project. The team has a small dataset and limited compute resources. What should the QA lead prioritize to ensure model validity?

   • AUse a large holdout test set to evaluate the final model
   • BImplement cross-validation and document the evaluation methodology
   • CSkip formal testing to save compute resources
   • DRely on the training accuracy as the primary performance metric
   Reveal Solution Discussion Select an answer first
4. 18foundation · easy

   What is a 'model version' in the context of AI/ML development?

   • AThe number of training epochs used to train the model
   • BThe name of the algorithm used in the model
   • CA snapshot of the model's code, data, and parameters at a specific point in time
   • DThe accuracy of the model on a validation set
   Reveal Solution Discussion Select an answer first
5. 19foundation · easy

   Which activity is part of a QA procedure for an AI/ML model?

   • ARunning a reproducibility check to confirm that the same training run produces the same results
   • BSelecting the cloud provider for model hosting
   • CDefining the model's business objectives
   • DWriting user documentation for the model's API
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   Why is it important to log and tag model versions and hyperparameters?

   • ATo provide traceability so that any model result can be linked to its exact configuration
   • BTo increase the model's accuracy on the test set
   • CTo automatically deploy the model to production
   • DTo reduce the amount of training data required
   Reveal Solution Discussion Select an answer first