Aligning before Aggregating: Enabling Communication Efficient Cross-Domain Federated Learning via Consistent Feature Extraction

Keyphrases

• Feature Extraction 100%
• Federated Learning 100%
• Cross-domain 100%
• Communication Efficiency 100%
• Consistent Feature 100%
• Global Targets 60%
• Feature Space 40%
• Feature Extractor 40%
• Learning Methods 40%
• Feature Shift 40%
• Feature Discrimination 40%
• Learning Model 20%
• Communication Cost 20%
• Performance Degradation 20%
• Contrastive Learning 20%
• Uniformly Distributed 20%
• Convergence Analysis 20%
• Hypersphere 20%
• Personalized Features 20%
• Lightweight Classifier 20%
• Domain Case 20%

Computer Science

• Feature Extraction 100%
• Federated machine learning 100%
• Feature Space 40%
• Communication Cost 20%
• Contrastive Learning 20%
• Extracted Feature 20%
• Mathematical Convergence 20%
• Performance Degradation 20%