Detecting, mitigating, and monitoring bias with SageMaker Clarify
SageMaker Clarify is a feature within the SageMaker service you can use for bias and explainability across your ML workflows. It has a nice integration with SageMaker’s Data Wrangler, a fully managed UI for tabular data analysis and exploration. This includes nearly 20 bias metrics, statistical terms you can study and use to get increasingly more precise about how your model interacts with humanity. I’ll spare you the mathematics here, but feel free to read more about them in my blog post on the topic here: https://towardsdatascience.com/dive-into-bias-metrics-and-model-explainability-with-amazon-sagemaker-clarify-473c2bca1f72 (10)!
Arguably more relevant for this book are Clarify’s vision and language features! This includes explaining image classification and object detection, along with language classification and regression. This should help you immediately understand what is driving your...
