Exam AWS Certified Machine Learning - Specialty topic 1 question 204 discussion

A is correct. "If the statistical nature of the data that your model receives while in production drifts away from the nature of the baseline data it was trained on, the model begins to lose accuracy in its predictions. Amazon SageMaker Model Monitor uses rules to detect data drift and alerts you when it happens." https://docs.aws.amazon.com/sagemaker/latest/dg/model-monitor-data-quality.html

A. YES - Model Monitor can validate distribution of input data B. NO - a model quality baseline is for model performance eg. precision, F1 score, etc. C. NO - Model Monitor is the right tool D. NO - Model Monitor is the right tool

it's a problem of monitoring data distributions.

The reason for this choice is that Amazon SageMaker Model Monitor is a feature of Amazon SageMaker that allows you to monitor and analyze your machine learning models in production. Model Monitor can automatically detect data drift and other data quality issues by comparing your live data with a baseline dataset that you provide1. Model Monitor can also emit metrics and alerts when it detects violations of the constraints that you define or that it suggests based on the baseline2.

A is correct. The best solution to meet the requirements is to use Amazon SageMaker Model Monitor to create a data quality baseline. The ML team can set up a data quality baseline to detect when the input data to the model has drifted significantly from the historical distribution of the data. When data drift occurs, the Model Monitor emits a metric that can trigger an alarm in Amazon CloudWatch. The ML team can use this alarm to investigate and take corrective action. Option B is incorrect because model quality baseline monitors model performance, not the input data quality. Option C is incorrect because Amazon SageMaker Debugger is used to debug machine learning models and to identify problems with model training, not data quality. Option D is incorrect because Amazon CloudWatch does not provide any features to monitor data drift in the input data used for the machine learning model.

Data monitor https://docs.aws.amazon.com/sagemaker/latest/dg/model-monitor-data-quality.html Properties of independent variables changes due to seasonality, customer preferences Model monitor https://docs.aws.amazon.com/sagemaker/latest/dg/model-monitor-model-quality.html Concept of what is spam email, changes over time "The company's ML team determines that the inaccuracies are occurring because of a change in the value distributions of the model features.” They know model features that is data for model input is changing so we monitor data https://pair-code.github.io/what-if-tool/learn/tutorials/features-overview/ a

Data monitor https://docs.aws.amazon.com/sagemaker/latest/dg/model-monitor-data-quality.html Properties of independent variables changes due to seasonality, customer preferences Model monitor https://docs.aws.amazon.com/sagemaker/latest/dg/model-monitor-model-quality.html Concept of what is spam email, changes over time "The company's ML team determines that the inaccuracies are occurring because of a change in the value distributions of the model features.” They know model features that is data for model input is changing so we monitor data https://pair-code.github.io/what-if-tool/learn/tutorials/features-overview/ a

I think the answer is B. Data quality can be monitored via model monitor model quality baseline.

B. Since it's "a change in the value distributions of the model features".

What is the difference of ans A and B?