AIGC-assisted Federated Learning for Edge Intelligence: Architecture Design, Research Challenges and Future Directions

TL;DR

The work tackles data heterogeneity in federated learning by introducing GenFL, an AIGC-assisted FL framework that leverages diffusion-based data augmentation to supplement clients' local data. GenFL operates with server-side AIGC, implementing a workflow of labels sharing, data augmentation, local training, and a weighted aggregation policy that blends FL updates with augmented data. Empirical results on CIFAR-10 and CIFAR-100 show GenFL achieving faster convergence and higher accuracy than baselines under non-IID distributions, demonstrating the potential to mitigate non-IID bottlenecks in edge learning. The study highlights open directions in improving data generation quality, designing adaptive aggregation strategies, incentivizing client participation, and developing resource-aware deployment to advance AIGC-enabled edge intelligence.

Abstract

Federated learning (FL) can fully leverage large-scale terminal data while ensuring privacy and security, and is considered as a distributed alternative for the centralized machine learning. However, the issue of data heterogeneity poses limitations on FL's performance. To address this challenge, artificial intelligence-generated content (AIGC) which is an innovative data synthesis technique emerges as one potential solution. In this article, we first provide an overview of the system architecture, performance metrics, and challenges associated with AIGC-assistant FL system design. We then propose the Generative federated learning (GenFL) architecture and present its workflow, including the design of aggregation and weight policy. Finally, using the CIFAR10 and CIFAR100 datasets, we employ diffusion models to generate dataset and improve FL performance. Experiments conducted under various non-independent and identically distributed (non-IID) data distributions demonstrate the effectiveness of GenFL on overcoming the bottlenecks in FL caused by data heterogeneity. Open research directions in the research of AIGC-assisted FL are also discussed.

Figures (6)

• Figure 1: The impacts of data distribution on the training performance.
• Figure 2: Examples of generated images.
• Figure 3: AIGC-assistant FL: architecture, metrics and challenges.
• Figure 4: Workflow of proposed AIGC-assistant FL architecture.
• Figure 5: Accuracy on CIFAR-10 with different Dirichlet distribution.