Breaking barriers in federated transfer learning: A systematic review on federated transfer learning for non-overlapping domains

Key Points

• Federated transfer learning enables models to be trained on decentralized data across non-overlapping domains, enhancing collaboration.
• Key evidence reveals improvements in predictive accuracy, indicating a better adaptation of models between distinct data sources.
• This systematic review evaluates various methodologies in federated transfer learning while focusing on non-overlapping domains, analyzing recent advancements.
• Findings may guide future research, highlighting the need for broader applicability to ensure effective learning across different domains.