AIncorrect
Amazon SageMaker Ground Truth
Ground Truth is a data labeling service that adds labels to training data.
It is not a capability for detecting drift after deployment, so this is incorrect.
A company wants to detect "drift", where the distribution of input data gradually shifts from training time, for a machine learning model deployed to production, and to catch quality degradation early. Which AWS capability is the MOST suitable?
1 / 1
Select an answer
CorrectB
Explanation
A question about choosing the service that detects drift after deployment.
- 1「detect "drift"」Capture the distribution shift from training time
- 2「catch quality degradation early」Continuous monitoring in production = SageMaker Model Monitor
BCorrect
Amazon SageMaker Model Monitor
Correct. SageMaker Model Monitor continuously monitors the input data and model quality of a production model and detects drift (distribution shift) from training time and raises alerts.
CIncorrect
Amazon SageMaker JumpStart
JumpStart is a model hub for using pretrained models and solutions right away.
It is not a capability for detecting drift after deployment, so this is incorrect.
DIncorrect
Amazon SageMaker Data Wrangler
Data Wrangler is a tool for data preprocessing and feature transformation through a GUI.
It is not a capability for detecting drift after deployment, so this is incorrect.
Key Takeaway
Note the correct answer, SageMaker Model Monitor.
- It continuously monitors the input data statistics and quality of a production model and detects drift (distribution shift) from training time and raises alerts.
- It helps decide when to retrain.
Ground Truth (labeling), JumpStart (model hub), and Data Wrangler (preprocessing) are all services involved in the pre-training phase, not capabilities for monitoring after deployment.