Dynamic federated semi-supervised learning with flexible unlabeled sample selection

Key Points

• Improved model accuracy is achieved through flexible unlabeled sample selection, enhancing learning efficiency.
• A 23% increase in performance is noted compared to traditional methods, utilizing more relevant unlabeled data.
• Dynamic federated semi-supervised learning was employed, allowing adaptability in sample selection methodologies.
• The findings highlight the potential for better resource utilization in machine learning applications, requiring further exploration.