![[2508.01669] Bridging Generalization Gap of Heterogeneous Federated Clients Using Generative Models](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2508.01669] Bridging Generalization Gap of Heterogeneous Federated Clients Using Generative Models
Summary
This paper presents a novel model-heterogeneous federated learning framework that enhances generalization performance for clients with diverse architectures by using generative models to create synthetic data.
Why It Matters
As federated learning becomes increasingly vital for privacy-preserving machine learning, addressing the challenges posed by heterogeneous client data is crucial. This research offers a new approach that not only improves model accuracy but also reduces communication costs, making it relevant for practitioners in the field.
Key Takeaways
- Proposes a model-heterogeneous framework for federated learning.
- Enhances generalization performance without parameter aggregation.
- Utilizes generative models to create synthetic data for training.
- Demonstrates improved accuracy and reduced communication costs.
- Addresses challenges of data heterogeneity among federated clients.
Computer Science > Machine Learning arXiv:2508.01669 (cs) [Submitted on 3 Aug 2025 (v1), last revised 13 Feb 2026 (this version, v2)] Title:Bridging Generalization Gap of Heterogeneous Federated Clients Using Generative Models Authors:Ziru Niu, Hai Dong, A.K. Qin View a PDF of the paper titled Bridging Generalization Gap of Heterogeneous Federated Clients Using Generative Models, by Ziru Niu and 2 other authors View PDF HTML (experimental) Abstract:Federated Learning (FL) is a privacy-preserving machine learning framework facilitating collaborative training across distributed clients. However, its performance is often compromised by data heterogeneity among participants, which can result in local models with limited generalization capability. Traditional model-homogeneous approaches address this issue primarily by regularizing local training procedures or dynamically adjusting client weights during aggregation. Nevertheless, these methods become unsuitable in scenarios involving clients with heterogeneous model architectures. In this paper, we propose a model-heterogeneous FL framework that enhances clients' generalization performance on unseen data without relying on parameter aggregation. Instead of model parameters, clients share feature distribution statistics (mean and covariance) with the server. Then each client trains a variational transposed convolutional neural network using Gaussian latent variables sampled from these distributions, and use it to generate synthe...
