A survey on federated learning: challenges and applications.

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Federated learning (FL) is a distributed machine learning approach enabling joint model training without centralizing sensitive data.
• Its inherent security and privacy advantages make it suitable for applications with strict data protection needs.

Purpose of the Study:

• To systematically review current research in federated learning.
• To identify and analyze key challenges and solutions in FL implementation.
• To provide insights into practical applications and future research directions.

Main Methods:

• Systematic literature review of federated learning research.
• Analysis of FL fundamentals, privacy/security mechanisms, communication overhead, and heterogeneity.
• Summary of FL's practical applications and future research trends.

Main Results:

• Federated learning faces significant bottlenecks in practical deployment, impacting model performance and efficiency.
• Key challenges include communication overhead and data heterogeneity.
• Ongoing research addresses these issues, leading to advancements in privacy-preserving AI.

Conclusions:

• Federated learning is a promising paradigm for privacy-preserving AI but requires addressing deployment challenges.
• Further research is needed to optimize FL efficiency and scalability for real-world applications.
• The review highlights FL's potential to drive intelligent development across various privacy-restricted domains.