Deployment, Monitoring, and Final Project

Five questions. Choose one answer for each, then submit to see your score and explanations.

The practice target is 80%. You can retry as often as you like, and every lesson stays accessible.

You can practice as a guest. Sign in before starting an attempt to save your scores.

1. What should accompany a deployed model for consistent predictions?A. The preprocessing and label mapping used by that model versionB. Only a screenshot of the training lossC. A new random normalization on every requestD. The final test labels in each prediction request

2. What is data drift?A. Any increase in the model filename lengthB. Proof that accuracy has improvedC. A requirement to retrain after every predictionD. A change in the distribution of incoming data

3. What makes a useful final project evaluation?A. Only training accuracy without a splitB. Repeated tuning on the final test setC. A defined task, a baseline, suitable metrics, and held-out evaluationD. Only a complex model name

4. A new model raises recall but produces more false alarms than the application can tolerate. What should you do before replacing the old model?A. Change test labels to match its predictionsB. Evaluate the trade-off against agreed requirements and keep a rollback optionC. Deploy automatically because one metric improvedD. Hide false positives from the report

5. Production images now come from a darker camera setup. What is a sound retraining approach?A. Review representative new examples, label them, and validate a candidate before replacing the current modelB. Train immediately on all unchecked predictions as if they were ground truthC. Overwrite the current model before measuring the new oneD. Assume more data always fixes every error

Review this module from its first lesson →

Deployment, Monitoring, and Final Project

Your result