IClientSelectionStrategy
Interfaces · AiDotNet.Interfaces
Selects which clients participate in a federated learning round.
Known implementations: ClientSelectionStrategyBase
How It Works
For Beginners: In federated learning, the server usually does not use every client in every round. Instead, it picks a subset of clients to reduce communication cost and handle device availability.
Different selection strategies optimize for different goals:
- Uniform: simple and fair
- Weighted: prefer clients with more data
- Stratified: ensure each group is represented
- Availability-aware: prefer clients likely to be online
- Performance-aware: prefer clients that historically help training
- Cluster-based: sample across diverse client behaviors
Methods
| Method | Summary |
|---|---|
List<int> SelectClients(ClientSelectionRequest request) |
Selects the clients to participate for the current round. |
string GetStrategyName() |
Gets the name of the selection strategy. |