Basic concepts of Amazon SageMaker Model Monitor and Amazon SageMaker Clarify
In this section, let's review the capabilities provided by two SageMaker features: Model Monitor and Clarify.
Amazon SageMaker Model Monitor provides capabilities to monitor data drift and the model quality of models deployed as SageMaker real-time endpoints. Amazon SageMaker Clarify provides capabilities to monitor the deployed model for bias and feature attribution drift. Using a combination of these two features, you can monitor the following four different aspects of ML models deployed on SageMaker:
- Data drift: If the live inference traffic data served by the deployed model is statistically different from the training data the model was trained on, the model prediction accuracy will start to deteriorate. Using a combination of a training data baseline and periodic monitoring to compare the incoming inference requests with the baseline data, SageMaker Model Monitor detects data drift...
