Interpretability

121 public types in this namespace, organized by kind.

Models & Types (93)

Type Summary
ALE2DResult<T> Represents the result of a 2D ALE analysis (feature interaction).
ALEResult<T> Represents the result of an ALE (Accumulated Local Effects) analysis.
AccumulatedLocalEffectsExplainer<T> Model-agnostic Accumulated Local Effects (ALE) explainer.
AnchorExplainer<T> Model-agnostic Anchor explainer that provides rule-based explanations.
AnchorExplanation<T> Represents an anchor explanation providing rule-based interpretations.
AttentionExplanation<T> Represents the result of an Attention Visualization analysis.
AttentionVisualizationExplainer<T> Attention Visualization explainer for Transformer models.
BackgroundSummary<T> Represents summarized background data for interpretability methods.
BasicFairnessEvaluator<T> Basic fairness evaluator that computes only fundamental fairness metrics.
BiasDetectionResult<T> Represents the results of a bias detection analysis.
ComprehensiveFairnessEvaluator<T> Comprehensive fairness evaluator that computes all major fairness metrics.
ConceptActivationVector<T> Represents a Concept Activation Vector.
ContrastiveExplainer<T> Contrastive explainer that answers “Why X and not Y?” questions.
ContrastiveExplanation<T> Represents the result of a Contrastive explanation.
CounterfactualExplainer<T> Model-agnostic Counterfactual explainer that finds minimal changes needed for a different prediction.
CounterfactualExplanation<T> Represents a counterfactual explanation showing minimal changes needed for a different outcome.
DeepLIFTExplainer<T> DeepLIFT (Deep Learning Important FeaTures) explainer for neural networks.
DeepLIFTExplanation<T> Represents the result of a DeepLIFT analysis.
DeepSHAPExplainer<T> DeepSHAP explainer combining GradientSHAP with DeepLIFT for efficient neural network explanations.
DeepSHAPExplanation<T> Represents the result of a DeepSHAP analysis.
DemographicParityBiasDetector<T> Detects bias using Demographic Parity (Statistical Parity Difference).
DiCEExplainer<T> DiCE (Diverse Counterfactual Explanations) explainer using genetic algorithm-based search.
DiCEExplanation<T> DiCE explanation containing multiple diverse counterfactuals.
DisparateImpactBiasDetector<T> Detects bias using the Disparate Impact metric (80% rule).
EqualOpportunityBiasDetector<T> Detects bias using Equal Opportunity metric (True Positive Rate difference).
ExplanationFaithfulnessReport<T> The auto-audit of how faithful a model’s explanations are — whether the features they highlight actually drive the model’s output.
FairnessMetrics<T> Represents fairness metrics for model evaluation.
FaithfulnessAuditor<T> Audits how faithful a per-feature attribution vector is to a model, using perturbation curves.
FaithfulnessReport<T> The faithfulness of a feature-attribution explanation to a model: do the features it highlights actually drive the model’s output? Reported as deletion/insertion AUC and ERASER-style comprehensiveness/sufficiency, plus a single normalized s…
FeatureAblationExplainer<T> Feature Ablation explainer for understanding feature importance by removal.
FeatureAblationExplanation<T> Result of feature ablation for a single input.
FeatureAttribution<T> Generic feature attribution result.
FeatureChange<T> Represents a single feature change in a counterfactual.
FeatureImportanceResult<T> Represents the result of permutation feature importance calculation.
FeatureInteractionExplainer<T> Model-agnostic Feature Interaction detector using Friedman’s H-statistic.
FeatureInteractionResult<T> Represents the result of a Feature Interaction analysis.
GlobalDeepSHAPExplanation<T> Represents global DeepSHAP feature importance.
GlobalFeatureAblationResult<T> Global feature ablation result across a dataset.
GlobalSHAPExplanation<T> Represents global SHAP explanations aggregated across multiple instances.
GlobalSurrogateExplainer<T> Global Surrogate Model explainer that approximates a complex model with an interpretable one.
GradCAMExplainer<T> Gradient-weighted Class Activation Mapping (Grad-CAM) explainer for CNNs.
GradCAMExplanation<T> Represents the result of a Grad-CAM analysis.
GradientSHAPExplainer<T> GradientSHAP explainer - a faster approximation of SHAP using gradients.
GradientSHAPExplanation<T> Represents the result of a GradientSHAP analysis.
GroupFairnessEvaluator<T> Group-level fairness evaluator that focuses on equalized performance across groups.
GuidedBackpropExplainer<T> Guided Backpropagation explainer for neural network visualization.
GuidedBackpropExplanation<T> Result of Guided Backpropagation explanation.
GuidedGradCAMExplainer<T> Guided GradCAM explainer combining GuidedBackprop with GradCAM for high-resolution explanations.
GuidedGradCAMExplanation<T> Guided GradCAM explanation result.
InfluenceFunctionExplainer<T> Influence Function explainer for training data attribution.
InfluenceFunctionResult<T> Result of influence function computation.
InputXGradientExplainer<T> Input × Gradient attribution explainer - multiplies input values by their gradients.
InputXGradientExplanation<T> Result of Input × Gradient attribution.
IntegratedGradientsExplainer<T> Integrated Gradients explainer for neural networks with gradient access.
IntegratedGradientsExplanation<T> Represents the result of an Integrated Gradients analysis.
LIMEExplainer<T> Model-agnostic LIME (Local Interpretable Model-agnostic Explanations) explainer.
LIMEExplanationResult<T> Represents a LIME explanation result.
LRPExplanation<T> Represents the result of an LRP analysis.
LayerAttributionExplainer<T> Layer-level attribution explainer for computing attributions at intermediate layers.
LayerAttributionResult<T> Result of layer attribution.
LayerGradCAMExplainer<T> Layer GradCAM (Gradient-weighted Class Activation Mapping) explainer.
LayerGradCAMExplanation<T> GradCAM explanation result.
LayerwiseRelevancePropagationExplainer<T> Layer-wise Relevance Propagation (LRP) explainer for neural networks.
LimeExplanation<T> Represents a LIME (Local Interpretable Model-agnostic Explanations) explanation for a prediction.
LocalTCAVExplanation<T> Represents a local TCAV explanation for a single input.
NeuronAttributionExplainer<T> Neuron-level attribution explainer for understanding individual neuron contributions.
NeuronAttributionResult<T> Result of neuron attribution.
NoiseTunnelExplainer<T, TExplanation> Noise Tunnel wrapper that smooths attributions by averaging over noisy inputs.
OcclusionExplainer<T> Occlusion explainer for image and sequential data interpretation.
OcclusionExplanation<T> Result of occlusion analysis.
PartialDependence2DResult<T> Represents the result of a 2D partial dependence analysis (feature interaction).
PartialDependenceData<T> Represents partial dependence data showing how features affect predictions.
PartialDependenceExplainer<T> Model-agnostic Partial Dependence Plot (PDP) explainer with Individual Conditional Expectation (ICE) curves.
PartialDependenceResult<T> Represents the result of a partial dependence analysis.
PermutationFeatureImportance<T> Model-agnostic Permutation Feature Importance calculator.
PertinentFeature<T> Represents a pertinent feature in a contrastive explanation.
PrototypeExplainer<T> Prototype-based explainer that explains predictions using similar examples.
PrototypeExplanation<T> Represents the result of a Prototype-based explanation.
PrototypeMatch<T> Represents a matched prototype.
SHAPExplainer<T> Model-agnostic SHAP (SHapley Additive exPlanations) explainer using Kernel SHAP algorithm.
SHAPExplanation<T> Represents a SHAP explanation for a single prediction.
SaliencyMapExplainer<T> Saliency Map explainer using gradient-based methods.
SaliencyMapExplanation<T> Represents the result of a Saliency Map analysis.
SelfInfluenceResult<T> Result of self-influence computation.
SingleCounterfactual<T> Represents a single counterfactual explanation.
SmartExplainerSelector<T> Automatically selects the optimal explainer based on model type and provides caching for batch explanations.
SurrogateExplanation<T> Represents a global surrogate model explanation.
TCAVExplainer<T> Testing with Concept Activation Vectors (TCAV) explainer.
TCAVResult<T> Represents the result of a TCAV analysis for a single concept.
TCAVResults<T> Represents TCAV results for multiple concepts.
TracInResult<T> Result of TracIn computation.
TreeSHAPExplainer<T> TreeSHAP explainer for computing exact SHAP values for tree-based models.
TreeSHAPExplanation<T> Represents the result of a TreeSHAP analysis.

Base Classes (2)

Type Summary
BiasDetectorBase<T> Base class for all bias detectors that identify unfair treatment in model predictions.
FairnessEvaluatorBase<T> Base class for all fairness evaluators that measure equitable treatment in models.

Interfaces (5)

Type Summary
IConvolutionalNetwork<T, TInput, TOutput> Interface for convolutional neural networks that support Grad-CAM explanation.
IFeatureAttribution<T> A common accessor for the per-feature attribution vector an explanation carries, regardless of the method that produced it (SHAP values, integrated gradients, DeepLIFT contributions, ablation deltas, saliency, …).
IGPUAcceleratedExplainer<T> Interface for explainers that support GPU acceleration.
IGlobalAttributionExplainer<T> 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).
ITransformerNetwork<T, TInput, TOutput> Interface for transformer networks that support attention visualization.

Enums (14)

Type Summary
DeepLIFTRule DeepLIFT attribution rules.
DistanceMetric Distance metrics for prototype matching.
ExplainableModelType Model types that the smart selector recognizes for choosing the optimal explainer.
ExplainerType Explainer types available.
FairnessMetric Enumeration of fairness metrics for model evaluation.
FeatureType Type of feature for counterfactual constraints.
InterpretationMethod Enumeration of interpretation methods supported by interpretable models.
InverseHessianMethod Methods for computing inverse Hessian-vector products.
LRPRule LRP propagation rules.
LayerAttributionMethod Layer attribution methods.
NeuronAttributionMethod Neuron attribution methods.
NoiseTunnelType Types of noise tunnel aggregation methods.
OcclusionShape Shape of occlusion patches.
SaliencyMethod Saliency computation methods.

Options & Configuration (1)

Type Summary
SmartExplainerOptions Options for smart explainer selector.

Helpers & Utilities (6)

Type Summary
BackgroundSummarizer<T> Provides methods for summarizing background data for SHAP and other interpretability methods.
GPUExplainerHelper<T> Provides GPU acceleration for interpretability explainers.
InputGradientHelper<T> Provides unified input gradient computation for interpretability explainers.
InterpretabilityMetricsHelper<T> Provides static utility methods for computing interpretability and fairness metrics.
InterpretableModelHelper Provides static helper methods for model interpretability operations using production-ready explainer algorithms.
NoiseTunnelFactory Convenience factory for creating NoiseTunnel wrappers for common explainer types.