Struct LearningService 

pub struct LearningService { /* private fields */ }

Main learning service for GNN-based embedding refinement.

This service manages:

• GNN model training on transition graphs
• Embedding refinement through message passing
• Edge prediction for relationship modeling
• Continual learning with EWC regularization

Implementations

impl LearningService

pub fn new(config: LearningConfig) -> Self

Create a new learning service with the given configuration

pub fn with_repository(self, repository: Arc<dyn LearningRepository>) -> Self

Create a learning service with a repository

pub fn config(&self) -> &LearningConfig

Get the current configuration

pub fn model_type(&self) -> GnnModelType

Get the model type

pub async fn start_session(&self) -> LearningResult<String>

Start a new training session

pub async fn train_epoch( &self, graph: &TransitionGraph, ) -> LearningResult<TrainingMetrics>

Train a single epoch on the transition graph

Arguments

• graph - The transition graph to train on

Returns

Training metrics for the epoch

pub async fn refine_embeddings( &self, embeddings: &[(EmbeddingId, Vec<f32>)], ) -> LearningResult<Vec<RefinedEmbedding>>

Refine embeddings using the trained GNN model

Arguments

• embeddings - Input embeddings to refine

Returns

Refined embeddings with quality scores

pub async fn predict_edge( &self, from: &[f32], to: &[f32], ) -> LearningResult<f32>

Predict edge weight between two embeddings

Arguments

• from - Source embedding
• to - Target embedding

Returns

Predicted edge weight (0.0 to 1.0)

pub async fn complete_session(&self) -> LearningResult<()>

Complete the current training session

pub async fn fail_session(&self, error: impl Into<String>) -> LearningResult<()>

Fail the current session with an error

pub async fn get_session(&self) -> Option<LearningSession>

Get the current session status

pub async fn consolidate_ewc( &self, graph: &TransitionGraph, ) -> LearningResult<()>

Save EWC state from current model for future regularization