Abstract:Recently, semi-supervised federated learning (semi-FL) has been proposed to handle the commonly seen real-world scenarios with labeled data on the server and unlabeled data on the clients. However, existing methods face several challenges such as communication costs, data heterogeneity, and training pressure on client devices. To address these challenges, we introduce the powerful diffusion models (DM) into semi-FL and propose FedDISC, a Federated Diffusion-Inspired Semi-supervised Co-training method. Specifically, we first extract prototypes of the labeled server data and use these prototypes to predict pseudo-labels of the client data. For each category, we compute the cluster centroids and domain-specific representations to signify the semantic and stylistic information of their distributions. After adding noise, these representations are sent back to the server, which uses the pre-trained DM to generate synthetic datasets complying with the client distributions and train a global model on it. With the assistance of vast knowledge within DM, the synthetic datasets have comparable quality and diversity to the client images, subsequently enabling the training of global models that achieve performance equivalent to or even surpassing the ceiling of supervised centralized training. FedDISC works within one communication round, does not require any local training, and involves very minimal information uploading, greatly enhancing its practicality. Extensive experiments on three large-scale datasets demonstrate that FedDISC effectively addresses the semi-FL problem on non-IID clients and outperforms the compared SOTA methods. Sufficient visualization experiments also illustrate that the synthetic dataset generated by FedDISC exhibits comparable diversity and quality to the original client dataset, with a neglectable possibility of leaking privacy-sensitive information of the clients.

Exploring One-Shot Semi-supervised Federated Learning with Pre-trained Diffusion Models

Exploring One-shot Semi-supervised Federated Learning with A Pre-trained Diffusion Model

FedDGP: Disentangling Global and Personal Models for Federated Learning

FedDEO: Description-Enhanced One-Shot Federated Learning with Diffusion Models

SemiDFL: A Semi-Supervised Paradigm for Decentralized Federated Learning

Semi-Supervised Federated Learning with Non-Iid Data: Algorithm and System Design

One-Shot Federated Learning with Classifier-Guided Diffusion Models

A Practical Data-Free Approach to One-shot Federated Learning with Heterogeneity

One-shot Federated Learning via Synthetic Distiller-Distillate Communication

FedDiff: Diffusion Model Driven Federated Learning for Multi-Modal and Multi-Clients

Clients Help Clients: Alternating Collaboration for Semi-Supervised Federated Learning

Semi-HFL: semi-supervised federated learning for heterogeneous devices

Navigating Data Heterogeneity in Federated Learning A Semi-Supervised Federated Object Detection

FedBiP: Heterogeneous One-Shot Federated Learning with Personalized Latent Diffusion Models

Efficient Semi-Supervised Federated Learning for Heterogeneous Participants

Federated Semi-Supervised Learning with Class Distribution Mismatch

FedDistill: Global Model Distillation for Local Model De-Biasing in Non-IID Federated Learning

SemiSFL: Split Federated Learning on Unlabeled and Non-IID Data

FedDM: Enhancing Communication Efficiency and Handling Data Heterogeneity in Federated Diffusion Models

Disentangling data distribution for Federated Learning

FedDC: Federated Learning with Non-IID Data via Local Drift Decoupling and Correction