IFederatedGraphTrainer<T>

Interfaces · AiDotNet.FederatedLearning.Graph

Orchestrates federated learning across clients holding subgraphs of a larger graph.

Known implementations: SubgraphFederatedTrainer<T>

For Beginners

Standard FL assumes clients have independent datasets. Graph FL is different because clients’ data is interconnected — edges may cross client boundaries. This trainer handles the unique challenges of graph FL:

Properties

Property Summary
int ClientCount { get; } Gets the number of registered clients.

Methods

Method Summary
Tensor<T> ExecuteRound(int roundNumber) Executes one round of federated graph learning.
Tensor<T> GetGlobalModel() Gets the current global GNN model parameters.
void DiscoverCrossClientEdges() Initiates cross-client edge discovery between all client pairs.
void RegisterSubgraph(int clientId, Tensor<T> adjacency, Tensor<T> nodeFeatures, Tensor<T> nodeLabels) Registers a client’s subgraph with the trainer.