A multi-head heterogeneous federated learning system with gradient- and data-based head embeddings plus nearest-neighbor selection reports 24.9% to 94.1% error reduction over single-source baselines on power-consumption prediction.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Heterogeneous Federated Learning Systems for Time-Series Power Consumption Prediction with Multi-Head Embedding Mechanism
A multi-head heterogeneous federated learning system with gradient- and data-based head embeddings plus nearest-neighbor selection reports 24.9% to 94.1% error reduction over single-source baselines on power-consumption prediction.