IGlobalAttributionExplainer<T>

Interfaces · AiDotNet.Interpretability

An opt-in capability for explainers that can produce a single global per-feature attribution vector over a dataset (typically the mean absolute local attribution).

Known implementations: DeepLIFTExplainer<T>, DeepSHAPExplainer<T>, FeatureAblationExplainer<T>, GradientSHAPExplainer<T>, InputXGradientExplainer<T>, IntegratedGradientsExplainer<T>

How It Works

The explainer implements this using its own (concrete) explanation type, so callers get a uniform attribution vector without dealing with the open explanation-type generic on the local/global explainer interfaces.

Methods

Method Summary
Vector<T> ComputeGlobalAttributions(Matrix<T> data) Computes a global per-feature attribution vector over the supplied data.