PMLE Monitoring ML Solutions Practice Question

An ML engineer is configuring Vertex AI Model Monitoring for a deployed model that receives a mix of numerical and categorical features. The engineer wants to ensure that the monitoring job can detect both data drift and training-serving skew. Which two configurations are required to enable both types of detection? (Choose two.)

⚠ Common exam trap

The trap here is conflating drift and skew detection configurations, assuming that one baseline or objective can enable both, when each requires its own setup.

Answer choices

Why each option matters

Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.

Correct answer & explanation

✓

Enable skew detection and specify the training dataset used to train the model.

To detect both data drift and training-serving skew in Vertex AI Model Monitoring, you must enable each detection type and provide the appropriate baseline. Skew detection requires the training dataset to compare against serving data. Drift detection requires a baseline dataset, often a sample of recent serving data. These two configurations are independent and both necessary to achieve the desired monitoring.

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

✓
Enable skew detection and specify the training dataset used to train the model.
Why this is correct
Training-serving skew detection requires comparing the feature distributions of the training data to the serving data. Therefore, you must enable skew detection and provide the training dataset as the baseline. This allows Vertex AI Model Monitoring to compute the training distribution and detect any significant differences in serving data, which is essential for identifying skew.
✗
Set the monitoring objective to 'drift' and provide a training dataset for baseline.
Why it's wrong here
Setting the objective to 'drift' enables drift detection, which compares serving data to a baseline. However, to also detect training-serving skew, you need to enable skew detection separately, which requires a training dataset to compute the training distribution. Just setting the objective to 'drift' does not enable skew detection. This configuration alone is insufficient.
✗
Set the monitoring frequency to at least once per hour to capture both types of drift.
Why it's wrong here
Monitoring frequency determines how often the job runs, but it does not enable drift or skew detection. While a higher frequency can provide more timely alerts, it is not a requirement for enabling the detection types. Both drift and skew can be detected at any frequency. This option does not contribute to enabling the detection capabilities.
✓
Enable drift detection and specify a baseline dataset (e.g., a sample of recent serving data).
Why this is correct
Data drift detection compares the current serving data to a baseline distribution, which can be a sample of recent serving data or another reference. By enabling drift detection and providing a baseline dataset, the monitoring job can detect changes in the input feature distributions over time. This is separate from skew detection, which compares to training data.
✗
Configure alerting via Cloud Monitoring to notify when either drift or skew exceeds thresholds.
Why it's wrong here
Alerting is important for acting on detected drift or skew, but it does not enable the detection itself. The monitoring job must be configured to compute drift and skew metrics first. Cloud Monitoring alerts are set up based on those metrics, but they are not a prerequisite for detection. This option is about notification, not detection configuration.

Go deeper

PMLE Monitoring ML Solutions Practice Question

Why each option matters

Option-by-option breakdown

Related to this question