Quiz: Ablation Evidence: Measuring What the Formal Layer Contributes

Ablation testing measures the individual contributions of components in a machine learning system by removing or disabling them and comparing the results to the original system.

In clinical decision support systems, ablation testing can be used to determine the impact of formal models such as ontologies on clinical recommendations.

Feature removal is an approach to ablation testing that involves disabling specific components of the formal layer such as particular ontology classes or relationships.

The purpose of measuring impact in ablation testing is to quantify how much outputs change when formal layer components are removed from the system.

Organisations implementing ablation testing should document which components were removed, what measurements were taken, and how outputs changed to support verification claims.