Collaborative optimization enhances training efficiency and reduces loss in resource-constrained environments, suggesting improvements for predictive maintenance and anomaly detection.
Key Points
The proposed method enhances training efficiency and reduces model loss by 41.7% compared to traditional methods.
It integrates client-side feature learning with hierarchical federated aggregation for better performance.
Leveraging the Kepler Optimization Algorithm, it optimizes model structure and scheduling strategies effectively.
This framework offers a scalable solution for training large models in resource-constrained industrial settings.