InfluenceFunctionResult<T>

Models & Types · AiDotNet.Interpretability.Explainers

Result of influence function computation.

Constructors

Constructor Summary
InfluenceFunctionResult(Vector<T> testInput, T testLabel, Vector<T> influences, Vector<T> prediction, T loss) Initializes a new influence function result.

Properties

Property Summary
T Loss { get; } Gets the loss on the test input.
T TestLabel { get; } Gets the test label.
Vector<T> Influences { get; } Gets influence scores for each training sample.
Vector<T> Prediction { get; } Gets the model’s prediction on the test input.
Vector<T> TestInput { get; } Gets the test input that was explained.
int NumTrainingSamples { get; } Gets the number of training samples.

Methods

Method Summary
IEnumerable<ValueTuple<int, T>> GetMostHarmful(int k) Gets the most harmful training samples (lowest/most negative influence).
IEnumerable<ValueTuple<int, T>> GetMostHelpful(int k) Gets the most helpful training samples (highest positive influence).
IEnumerable<ValueTuple<int, T>> GetTopInfluential(int k) Gets the top K most influential training samples.
string ToString() Returns a human-readable summary.