This is a dedicated watch page for a single video.
Concept: Fairlearn is a Python package designed to analyze machine learning models and assess disparities in predictions and prediction performance across sensitive features. It integrates seamlessly with Azure Machine Learning, allowing users to conduct experiments and upload dashboard metrics to their Azure workspace. The appropriate parity constraint depends on the mitigation technique and the specific fairness criteria being applied. Question: Which Fairlearn constraint ensures that, when used with reduction-based mitigation algorithms such as Exponentiated Gradient and Grid Search, the error for each sensitive feature group remains within a specified threshold of the overall error rate?