ShapleyValueEvaluator<T>
Models & Types · AiDotNet.FederatedLearning.Fairness
Exact Shapley value evaluator: computes each client’s marginal contribution across all coalitions.
Inheritance: FederatedLearningComponentBase<T> → ShapleyValueEvaluator<T>
Implements: IClientContributionEvaluator<T>
For Beginners
The Shapley value (from game theory, Nobel Prize 2012) is the fairest way to divide credit among participants. It answers: “On average, how much does each client improve the model when added to any possible subset of other clients?”
How It Works
How it works:
Warning: Exact Shapley requires evaluating 2^N coalitions, making it impractical for
more than ~15 clients. Use DataShapleyEvaluator for larger federations.
Constructors
| Constructor | Summary |
|---|---|
ShapleyValueEvaluator(ContributionEvaluationOptions options) |
Initializes a new instance of ShapleyValueEvaluator. |
Properties
| Property | Summary |
|---|---|
string MethodName { get; } |
Methods
| Method | Summary |
|---|---|
Dictionary<int, double> EvaluateContributions(Dictionary<int, Tensor<T>> clientModels, Tensor<T> globalModel, Dictionary<int, List<Tensor<T>>> clientHistories) |
|
HashSet<int> IdentifyFreeRiders(Dictionary<int, double> contributionScores) |