A Review on Federated Learning with Dual Knowledge Transfer
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
With smartphones and edge devices being commonplace, expansive personal photo data has become incredibly common. As we embark on applications that utilize personal photo data to train effective classification models, including smart photo organization, the risks of privacy and the complexities of deploying centralized training weigh heavily on us. With the capabilities of Federated Learning (FL), users can collaboratively train models with other users without the sharing of raw personal data. However, FL is challenged by non-IID (non-independently and identically distributed) data, as dataset uniqueness stemmed from clients' unique photo collections is problematic because it yields non-IID client datasets that are often unbalanced too. FL methods, such as FedAvg, traditionally develop an adaptable global model while providing poorly personalized local models to users; newer approaches, like FedRep, improve personalization, but can instead worsen the performance of the global model. This paper presents an innovative dual knowledge transfer framework that mitigates the impact on i) local model personalization, and ii) global model generalization. The framework combines synthetic feature generation and bidirectional knowledge distillation to ensure users receive a personalized model while the users collaboratively improve the performance of the global model, all without compromising an individual's privacy.
Fields of Study & Subjects
Privacy-Preserving Technologies in Data
Domain Adaptation and Few-Shot Learning
Stochastic Gradient Optimization Techniques
Publication details
- Journal / source
- International Conference on Intelligent Systems for Communication, IoT and Security (ICISCoIS)
- Pages
- 1–6
- Publication date
- 8 January 2026
- Language
- EN