摘要
联邦学习是一种新兴的人工智能基础技术,其设计目标是在保障大数据共享交换时的信息安全、保护终端个人数据隐私、保证合法合规的前提下,在多参与者或多计算节点之间开展高效率的机器学习,有望成为下一代人工智能协同算法和协作网络的基础。对近期联邦学习的相关研究与成果进行了综合评述,并对未来发展趋势进行了展望。首先,从数据孤岛和隐私保护等两个角度介绍联邦学习的兴起背景,并概述联邦学习内涵与机理;随后,聚焦技术革新和标准化建设“双轮驱动”,概述联邦学习最新进展,并以公共文化服务领域为示范应用场景,介绍基于联邦学习技术的数据共享交换典型应用案例;最后,从通信费用高、统计异构性、系统异构性、隐私问题等四个角度讨论联邦学习面临的困难与挑战。
Federated learning is a new basic technology of artificial intelligence(AI).It aims at operating efficient machine learning among multi participants or multi computing nodes,on the premise of ensuring the information security when sharing and exchanging big data,protecting the personal data privacy of the terminal,and ensuring the legal compliance.It is expected to become the basis of the next generation of AI collaborative algorithms and collaborative networks.This paper reviews the recent research and achievements of Federated learning,and looks forward to the future development trend.First of all,this paper introduces the background of federal learning from the perspectives of data island and privacy protection,and summarizes its connotation and mechanism.Then,this paper summarizes the latest progress of federal learning from the perspectives of technological innovation and standardized construction.Moreover,take the public-cultural service field as the demonstration application,this paper introduces a typical application case of data sharing and exchange based on federal learning.Finally,the problems faced by federal learning about high communication costs,statistical heterogeneity,system heterogeneity and privacy issues,are discussed.
作者
王亚珅
WANG Yashen(China Academy of Electronics and Information Technology,National Engineering Laboratory for Public Safety Risk Perception and Control by Big Data(PSRPC),Beijing 100041,China)
出处
《无人系统技术》
2019年第6期58-62,共5页
Unmanned Systems Technology
基金
国家重点研发计划项目（2017YFC0820503）
中国电科新一代人工智能专项行动计划项目（AI20191125008）
中国电科装备预研联合基金项目（6141B08010102）
全国一体化国家大数据中心先导工程（17111001,17111002）
国家自然科学基金项目（U19B2026）
中国博士后科学基金第64批面上资助项目（2018M641436）
文旅部2018年度文化和旅游智库项目（18K01）
关键词
联邦学习
数据共享交换
人工智能
隐私安全
数据孤岛
机器学习
系统架构设计
分布式计算
Federated Learning
Data Sharing and Exchange
Artificial Intelligence
Privacy Security
Data Is⁃land
Machine Learning
System Architecture Design
Distributed Computing
