FLBench
收藏资源简介:
FLBench是一个针对联邦学习的基准套件,涵盖医疗、金融和AIoT三个领域。该数据集包含四个子数据集,分别来自医疗领域的ADNI和MIMIC-III、金融领域的Adult dataset以及AIoT领域的iNaturalist-User-120k。FLBench旨在评估联邦学习系统与算法的多个关键方面,如通信效率、场景转换、隐私保护、数据分布异质性和合作策略。通过提供可配置的场景,FLBench支持开发新型联邦学习算法,并已作为自动化部署工具开源,适用于多种平台。
FLBench is a benchmark suite for federated learning, covering three domains: healthcare, finance, and AIoT. It includes four sub-datasets, namely ADNI and MIMIC-III from the healthcare domain, the Adult dataset from the finance domain, and iNaturalist-User-120k from the AIoT domain. FLBench aims to evaluate multiple key aspects of federated learning systems and algorithms, such as communication efficiency, scenario transfer, privacy protection, data distribution heterogeneity, and collaboration strategies. By providing configurable scenarios, FLBench supports the development of novel federated learning algorithms, and it has been open-sourced as an automated deployment tool compatible with multiple platforms.




