You must measure whether a training dataset is imbalanced across a sensitive attribute before training. Which tool fits?
Why C is the answer
Amazon SageMaker Clarify
Clarify provides bias metrics before and after training, and supports explainability.
In detail
- Clarify measures bias in data before training.
- AWS closed SageMaker Clarify to new customers. Existing customers can keep using it.
- Clarify also reports bias after training.
Use case
An illustrative scenario, not a real incident.
A lender checks whether loan data under-represents one age group before training. Clarify reports bias metrics, so the team can correct it first.
See it in different ways
flowchart LR A["Raw data in S3"] --> B["Clean and transform"] B --> C["Feature engineering"] C --> D["Check quality and bias"] D --> E["Training set"]
mindmap
root((SageMaker AI))
Data
Data Wrangler
Ground Truth
Clarify
Train
JumpStart
Automatic model tuning
Debugger
Deploy
Pipelines
Model Registry
Operate
Model Monitorflowchart LR S0["Load dataset"] --> S1["Choose sensitive attribute"] --> S2["Clarify bias report"] --> S3["Correct the data"]
Key point: Clarify measures bias before training.
Why the other options are wrong
- A. Amazon SageMaker Feature Store
- Feature Store holds features.
- B. Amazon SageMaker Ground Truth
- Ground Truth labels data.
- D. Amazon SageMaker Autopilot
- Autopilot builds models.