摘要
随着移动端数据量的快速增长,传统的集中式数据处理方法正在变得不可行。因此移动边缘计算技术被提出,然而计算节点在协作处理数据的过程中可能会产生数据隐私泄露的问题,这大大阻碍了该技术的发展。联邦学习技术提供了用于协作和安全的学习协议,可以作为移动边缘计算的操作系统,应用在移动边缘计算中安全地训练AI模型。NGBoost是一种新型且有效的数据处理方法,却不支持分布式环境。基于此,文章提出了面向移动边缘计算的联邦NGBoost处理方法,计算节点能够在保护数据隐私的前提下协作训练共同的NGBoost模型。实验表明,联邦NGBoost模型能够达到与集中式NGBoost模型近似的性能。
With the rapid growth of mobile data volume, traditional centralized data processing methods are becoming unfeasible. Therefore, Mobile Edge Computing Technique is proposed. However, data privacy leakage may occur in the process of collaborative data processing by computing nodes, which greatly hinders the development of this technology. Federated Learning Technology provides learning protocols for collaboration and security and can be used as an operating system for Mobile Edge Computing to safely train AI models in Mobile Edge Computing. NGBoost is a new and efficient data processing method that does not support distributed environments. Based on this, Federated NGBoost Technique for Mobile Edge Computing is proposed. Computing nodes can cooperatively train a common NGBoost Model under the premise of protecting data privacy. Experiments show that the Federated NGBoost Model can approximate the performance of the centralized NGBoost Model.
作者
徐倩
徐先栋
朱培康
Xu Qian;Xu Xiandong;Zhu Peikang(The 28th Research Institute of China Electronics Technology Group Corporation,Nanjing 210007,China;School of Telecommunications and Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210003)
出处
《信息化研究》
2021年第5期13-17,78,共6页
INFORMATIZATION RESEARCH
基金
南京邮电大学校科研基金项目(No.NY220007)。
