FederatedScope GNN：为联邦图学习提供统一、全面和高效的软件包

Title: FederatedScope-GNN: Towards a Unified, Comprehensive and Efficient  Package for Federated Graph Learning

Published: 2022-04-14

Url: http://arxiv.org/abs/2204.05562v3

Authors: Zhen Wang,Weirui Kuang,Yuexiang Xie,Liuyi Yao,Yaliang Li,Bolin Ding,Jingren Zhou

联邦学习（federated learning，FL）的惊人发展使计算机视觉和自然语言处理领域的各种任务受益匪浅，而TFF和FATE等现有框架使其在现实世界的应用中易于部署。然而，联邦图学习（federated graph learning，FGL）由于其独特的特性和要求，即使是思想图数据也很普遍，但并没有得到很好的支持。FGL相关框架的缺乏增加了实现可重复研究和在现实世界应用中部署的努力。基于这种强烈的需求，在本文中，我们首先讨论了创建一个易于使用的FGL包的挑战，并相应地提出了我们实现的包FederatedScope GNN（FS-G），它提供了（1）一个模块化和表达FGL算法的统一视图；（2） 全面的DataZoo和ModelZoo提供开箱即用的FGL功能；（3） 一个高效的模型自动调整组件；（4）现成的隐私攻击和防御能力。我们通过进行扩展实验来验证FS-G的有效性，这同时为社区获得了许多关于FGL的宝贵见解。此外，我们还使用FS-G为现实世界电子商务场景中的FGL应用程序提供服务，在这些场景中，所获得的改进表明了巨大的潜在商业利益。我们在https://github.com/alibaba/FederatedScope促进FGL的研究并实现广泛的应用。

The incredible development of federated learning (FL) has benefited varioustasks in the domains of computer vision and natural language processing, and the existing frameworks such as TFF and FATE has made the deployment easy inreal-world applications. However, federated graph learning (FGL), even though graph data are prevalent, has not been well supported due to its unique characteristics and requirements. The lack of FGL-related framework increasesthe efforts for accomplishing reproducible research and deploying in real-worldapplications. Motivated by such strong demand, in this paper, we first discussthe challenges in creating an easy-to-use FGL package and accordingly presentour implemented package FederatedScope-GNN (FS-G), which provides (1) a unified view for modularizing and expressing FGL algorithms; (2) comprehensive DataZooand ModelZoo for out-of-the-box FGL capability; (3) an efficient modelauto-tuning component; and (4) off-the-shelf privacy attack and defense abilities. We validate the effectiveness of FS-G by conducting extensiveexperiments, which simultaneously gains many valuable insights about FGL forthe community. Moreover, we employ FS-G to serve the FGL application inreal-world E-commerce scenarios, where the attained improvements indicate great potential business benefits. We publicly release FS-G, as submodules of FederatedScope, at https://github.com/alibaba/FederatedScope to promote FGL'sresearch and enable broad applications that would otherwise be infeasible due to the lack of a dedicated package.

整体架构如下：

框架已实现的方法：

基于消息传递的实现方案：

开源代码

https://github.com/alibaba/FederatedScope

FederatedScope 是一个综合性的联邦学习平台，为学术界和工业界的各种联邦学习任务提供方便的使用和灵活的定制。FederatedScope基于面向消息的框架，集成了丰富的功能集合，以满足联邦学习日益增长的需求，旨在构建一个易于使用的平台，以安全有效地促进学习。

Tutorial参见：https://federatedscope.io/

有完整的API：

涉及的已发表工作：

END