arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2406.02015cs.LGcs.DC
收藏

面向可复现研究的联邦持续学习参数化

Parameterizing Federated Continual Learning for Reproducible Research

Delft University of Technology(代尔夫特理工大学)

Bart Cox, Jeroen Galjaard, Aditya Shankar, Jérémie Decouchant, Lydia Y. Chen

AI总结:

针对联邦学习在异构环境和任务演变下的挑战,提出首个完全可配置的联邦持续学习框架 Freddie,通过容器化技术实现大规模部署,并验证了其在复杂场景下的有效性。

摘要:

联邦学习(FL)系统在异构且不断演变的环境中运行,这对其性能构成了挑战。在实际部署中,客户端的学习任务也会随时间推移而演变,这需要引入诸如持续学习等方法。为实现研究的可复现性,我们提出了一套实验最佳实践,能够精确捕捉并模拟复杂的学习场景。我们的框架 Freddie 是首个完全可配置的联邦持续学习(FCL)框架,得益于 Kubernetes 和容器化技术的使用,它可以无缝部署在大量机器上。我们在两个用例上展示了 Freddie 的有效性:(i) 在 CIFAR100 上的大规模 FL,以及 (ii) FCL 中的异构任务序列,这些用例突出了 FCL 场景中尚未解决的性能挑战。

英文摘要:

Federated Learning (FL) systems evolve in heterogeneous and ever-evolving environments that challenge their performance. Under real deployments, the learning tasks of clients can also evolve with time, which calls for the integration of methodologies such as Continual Learning. To enable research reproducibility, we propose a set of experimental best practices that precisely capture and emulate complex learning scenarios. Our framework, Freddie, is the first entirely configurable framework for Federated Continual Learning (FCL), and it can be seamlessly deployed on a large number of machines thanks to the use of Kubernetes and containerization. We demonstrate the effectiveness of Freddie on two use cases, (i) large-scale FL on CIFAR100 and (ii) heterogeneous task sequence on FCL, which highlight unaddressed performance challenges in FCL scenarios.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑