DeepLIFTExplanation<T>

Models & Types · AiDotNet.Interpretability.Explainers

Represents the result of a DeepLIFT analysis.

Implements: IFeatureAttribution<T>

Constructors

Constructor Summary
DeepLIFTExplanation()

Properties

Property Summary
DeepLIFTRule Rule { get; set; } Gets or sets the DeepLIFT rule used.
T BaselinePrediction { get; set; } Gets or sets the prediction at the baseline.
T CompletenessError { get; set; } Gets or sets the completeness error (difference between sum of attributions and delta output).
T DeltaOutput { get; set; } Gets or sets the difference between input and baseline predictions.
T InputPrediction { get; set; } Gets or sets the prediction at the input.
Vector<T> Attributions { get; set; } Gets or sets the feature attributions.
Vector<T> Baseline { get; set; } Gets or sets the baseline used.
Vector<T> Input { get; set; } Gets or sets the input instance.
int OutputIndex { get; set; } Gets or sets the output index that was explained.
string[] FeatureNames { get; set; } Gets or sets the feature names.

Methods

Method Summary
List<ValueTuple<string, T>> GetNegativeContributions() Gets negative contributions (features that decreased prediction).
List<ValueTuple<string, T>> GetPositiveContributions() Gets positive contributions (features that increased prediction).
List<ValueTuple<string, T>> GetSortedAttributions() Gets attributions sorted by absolute value (most important first).
Vector<T> GetFeatureAttributions()
string ToString() Returns a human-readable summary.