Federated Graph Learning (FGL) has emerged as a promising paradigm for breaking data silos in distributed private graphs data management. In practical scenarios involving complex and heterogeneous distributed graph data, personalized Federated Graph Learning (pFGL) aims to enhance model utility by training personalized models tailored to individual client needs, rather than relying on a universal global model. However, existing pFGL methods often require numerous communication rounds under heterogeneous client graphs, leading to significant security concerns and communication overhead. While One-shot Federated Learning (OFL) addresses these issues by enabling collaboration in a single round, existing OFL methods are designed for image-based tasks and ineffective for graph data, leaving a critical gap in the field. Additionally, personalized models often suffer from bias, failing to generalize effectively to minority data. To address these challenges, we propose the first one-shot personalized federated graph learning method for node classification, compatible with the Secure Aggregation protocol for privacy preservation. Specifically, for effective graph learning in a single communication round, our method estimates and aggregates class-wise feature distribution statistics to construct a global pseudo-graph on the server, facilitating the training of a global graph model. Moreover, to mitigate bias, we introduce a two-stage personalized training approach that adaptively balances local personal information and global insights from the pseudo-graph, improving both personalization and generalization. Extensive experiments conducted on 8 multi-scale graph datasets demonstrate that our method significantly outperforms state-of-the-art baselines across various settings.

翻译：联邦图学习已成为打破分布式私有图数据管理中数据孤岛的一种有前景的范式。在涉及复杂且异构的分布式图数据的实际场景中，个性化联邦图学习旨在通过训练适应个体客户端需求的个性化模型来提升模型效用，而非依赖一个通用的全局模型。然而，现有的个性化联邦图学习方法在异构客户端图数据下通常需要大量通信轮次，导致严重的安全隐患和通信开销。虽然单轮联邦学习通过实现单轮协作解决了这些问题，但现有的单轮联邦学习方法专为基于图像的任务设计，对图数据无效，这在该领域留下了一个关键空白。此外，个性化模型常受偏差影响，难以有效泛化到少数类数据。为应对这些挑战，我们提出了首个用于节点分类的单轮个性化联邦图学习方法，该方法兼容用于隐私保护的Secure Aggregation协议。具体而言，为实现单轮通信中的有效图学习，我们的方法估计并聚合类别的特征分布统计量，以在服务器上构建一个全局伪图，从而促进全局图模型的训练。此外，为减轻偏差，我们引入了一种两阶段个性化训练方法，该方法自适应地平衡来自伪图的本地个性化信息与全局洞察，从而同时提升个性化和泛化能力。在8个多尺度图数据集上进行的大量实验表明，我们的方法在各种设置下均显著优于最先进的基线模型。