IIncentiveMechanism<T>
Interfaces · AiDotNet.FederatedLearning.Fairness
Computes incentive rewards for federated learning participants based on their contributions.
Known implementations: ContributionBasedIncentive<T>
For Beginners
Clients need motivation to participate in federated learning. Without incentives, rational clients may free-ride (benefit from the global model without contributing their own data). An incentive mechanism assigns rewards proportional to each client’s contribution, encouraging high-quality participation.
How It Works
Example: In a data marketplace, a hospital contributing rare disease data (high marginal value) earns more than one contributing common cold data (low marginal value).
Properties
| Property | Summary |
|---|---|
string MechanismName { get; } |
Gets the name of this incentive mechanism. |
Methods
| Method | Summary |
|---|---|
Dictionary<int, double> ComputeRewards(Dictionary<int, double> contributionScores, double totalBudget) |
Computes incentive rewards for each client based on their contribution scores. |
Dictionary<int, double> ComputeTrustScores(Dictionary<int, List<double>> contributionHistory) |
Computes trust scores for clients based on their participation history. |