ContributionBasedIncentive<T>
Models & Types · AiDotNet.FederatedLearning.Fairness
Incentive mechanism that rewards clients proportional to their evaluated contribution.
Inheritance: FederatedLearningComponentBase<T> → ContributionBasedIncentive<T>
Implements: IIncentiveMechanism<T>
For Beginners
Clients need motivation to participate in federated learning. This mechanism distributes a reward budget proportionally: clients who contribute more valuable data get a bigger share. It also tracks trust over time — consistently helpful clients earn higher trust scores, while erratic or low-quality contributors are flagged.
How It Works
How rewards work:
Trust scoring: Trust is computed from contribution history using exponential moving average with consistency bonus. Clients with stable, positive contributions earn higher trust than those with sporadic or negative contributions.
Constructors
| Constructor | Summary |
|---|---|
ContributionBasedIncentive(ContributionEvaluationOptions options) |
Initializes a new instance of ContributionBasedIncentive. |
Properties
| Property | Summary |
|---|---|
string MechanismName { get; } |
Methods
| Method | Summary |
|---|---|
Dictionary<int, double> ComputeRewards(Dictionary<int, double> contributionScores, double totalBudget) |
|
Dictionary<int, double> ComputeTrustScores(Dictionary<int, List<double>> contributionHistory) |