联邦学习算法综述

doi:10.11959/j.issn.2096-0271.2020055

大数据 ›› 2020, Vol. 6 ›› Issue (6): 64-82.doi: 10.11959/j.issn.2096-0271.2020055

所属专题：联邦学习

联邦学习算法综述

王健宗¹,孔令炜¹,黄章成¹,陈霖捷¹,刘懿¹,何安珣¹,肖京²

¹ 平安科技（深圳）有限公司，广东深圳 518063
² 中国平安保险（集团）股份有限公司，广东深圳 518031

出版日期:2020-11-15 发布日期:2020-12-12
作者简介:王健宗（1983- ），男，博士，平安科技（深圳）有限公司副总工程师，资深人工智能总监，联邦学习技术部总经理。美国佛罗里达大学人工智能博士后，中国计算机学会（CCF）高级会员，CCF大数据专家委员会委员，曾任美国莱斯大学电子与计算机工程系研究员，主要研究方向为联邦学习和人工智能等|孔令炜（1995- ），男，平安科技（深圳）有限公司联邦学习团队算法工程师，CCF会员，主要研究方向为联邦学习系统和安全通信等|黄章成（1990- ），男，平安科技（深圳）有限公司联邦学习团队资深算法工程师，人工智能专家，CCF会员，主要研究方向为联邦学习、分布式计算及系统和加密通信等|陈霖捷（1994- ），男，平安科技（深圳）有限公司联邦学习团队算法工程师，主要研究方向为联邦学习与隐私保护、机器翻译等|刘懿（1994- ），女，平安科技（深圳）有限公司联邦学习团队算法工程师，主要研究方向为联邦学习系统等|何安珣（1990- ），女，平安科技（深圳）有限公司联邦学习团队高级算法工程师，CCF会员，主要研究方向为联邦学习技术在金融领域的落地应用、联邦学习框架搭建、加密算法研究和模型融合技术|肖京（1972- ），男，博士，中国平安保险（集团）股份有限公司首席科学家。2019年吴文俊人工智能科学技术奖杰出贡献奖获得者，CCF深圳会员活动中心副主席，主要研究方向为计算机图形学学科、自动驾驶、3D显示、医疗诊断、联邦学习等
基金资助:
国家重点研发计划基金资助项目(2018YFB1003503);国家重点研发计划基金资助项目(2018YFB0204400);国家重点研发计划基金资助项目(2017YFB1401202)

Research review of federated learning algorithms

Jianzong WANG¹,Lingwei KONG¹,Zhangcheng HUANG¹,Linjie CHEN¹,Yi LIU¹,Anxun HE¹,Jing XIAO²

¹ Ping An Technology (Shenzhen) Co.,Ltd.,Shenzhen 518063,China
² Ping An Insurance (Group) Company of China,Ltd.,Shenzhen 518031,China

Online:2020-11-15 Published:2020-12-12
Supported by:
The National Key Research and Development Program of China(2018YFB1003503);The National Key Research and Development Program of China(2018YFB0204400);The National Key Research and Development Program of China(2017YFB1401202)

摘要/Abstract

摘要：

近年来，联邦学习作为解决数据孤岛问题的技术被广泛关注，已经开始被应用于金融、医疗健康以及智慧城市等领域。从3个层面系统阐述联邦学习算法。首先通过联邦学习的定义、架构、分类以及与传统分布式学习的对比来阐述联邦学习的概念；然后基于机器学习和深度学习对目前各类联邦学习算法进行分类比较和深入分析；最后分别从通信成本、客户端选择、聚合方式优化的角度对联邦学习优化算法进行分类，总结了联邦学习的研究现状，并提出了联邦学习面临的通信、系统异构、数据异构三大难题和解决方案，以及对未来的期望。

关键词: 联邦学习, 算法优化, 大数据, 数据隐私

Abstract:

In recent years,federated learning has been proposed and received widespread attention to overcome data isolated island challenge.Federated learning related researches were adopted in areas such as financial field,healthcare domain and smart city related application.Federated learning concept was introduced into three different layers.The first layer introduced the definition,architecture,classification of federated learning and compared the federated learning with traditional distributed learning.The second layer presented comparison and analysis of federated learning algorithms from machine learning and deep learning aspects.The third layer separated federated learning optimization algorithms into three aspects to optimize federated learning algorithm through reducing communication cost,selecting proper clients and different aggregation method.Finally,the current research status and three main challenges on communication,heterogeneity of system and data to be solved were concluded,and the future prospects in federated learning domain were proposed.

Key words: federated learning, algorithm optimization, big data, data privacy

中图分类号:

TP311

王健宗, 孔令炜, 黄章成, 陈霖捷, 刘懿, 何安珣, 肖京. 联邦学习算法综述[J]. 大数据, 2020, 6(6): 64-82.

Jianzong WANG, Lingwei KONG, Zhangcheng HUANG, Linjie CHEN, Yi LIU, Anxun HE, Jing XIAO. Research review of federated learning algorithms[J]. Big Data Research, 2020, 6(6): 64-82.

图/表 4

图1

图2

表1

表2

参考文献 67

[1]	LECUN Y , BENGIO Y , HINTON G . Deep learning[J]. Nature, 2015,521(7553): 436-444.
[2]	王健宗, 黄章成, 肖京 . 人工智能赋能金融科技[J]. 大数据, 2018,4(3): 114-119.
	WANG J Z , HUANG Z C , XIAO J . Artificial intelligence energize Fintech[J]. Big Data Research, 2018,4(3): 114-119.
[3]	KAIROUZ P , MCMAHAN H B , AVENT B ,et al. Advances and open problems in federated learning[J]. arXiv preprint,2019,arXiv:1912.04977.
[4]	YANG Q , LIU Y , CHEN T ,et al. Federated machine learning:concept and applications[J]. ACM Transactions on Intelligent Systems and Technology, 2019,10(2): 1-19.
[5]	LI T , SAHU A K , TALWALKAR A ,et al. Federated learning:challenges,methods,and future directions[J]. IEEE Signal Processing Magazine, 2020,37(3): 50-60.
[6]	MEHMOOD A , NATGUNANATHAN I , XIANG Y ,et al. Protection of big data privacy[J]. IEEE Access, 2016,4: 1821-1834.
[7]	方滨兴, 贾焰, 李爱平 ,等. 大数据隐私保护技术综述[J]. 大数据, 2016,2(1): 1-18.
	FANG B X , JIA Y , LI A P ,et al. Privacy preservation in big data:a survey[J]. Big Data Research, 2016,2(1): 1-18.
[8]	KONE?NY J , MCMAHAN H B , RAMAGE D ,et al. Federated optimization:distributed machine learning for ondevice intelligence[J].. arXiv preprint,2016,arXiv:1610.02527,
[9]	KONE?NY J , MCMAHAN H B , YU F X ,et al. Federated learning:strategies for improving communication efficiency[J]. arXiv preprint,2016,arXiv:1610.05492.
[10]	MCMAHAN H B , MOORE E , RAMAGE D ,et al. Federated learning of deep networks using model averaging[J]. arXiv preprint,2016,arXiv:1602.05629.
[11]	MCMAHAN H B , MOORE E , RAMAGE D ,et al. Communication-efficient learning of deep networks from decentralized data[C]// Conference on Artificial Intelligence and Statistics.[S.l.:s.n]. 2017.
[12]	LI T , SANJABI M , BEIRAMI A ,et al. Fair resource allocation in federated learning[J]. arXiv preprint,2019,arXiv:1905.10497.
[13]	CHEN Y , SUN X Y , JIN Y C . Communication-efficient federated deep learning with layer wise asynchronous model update and temporally weighted aggregation[J]. IEEE Transactions on Neural Networks and Learning Systems,2019:Accepted.
[14]	REHAK D R , DODDS P , LANNOM L . A model and infrastructure for federated learning content repositories[C]// Interoperability of Web-Based Educational Systems Workshop.[S.l.:s.n]. 2005.
[15]	LI M , ANDERSEN D G , PARK J W ,et al. Scaling distributed machine learning with the parameter server[C]// The 11th USENIX Symposium on Operating Systems Design and Implementation.[S.l.:s.n]. 2014: 583-598.
[16]	LIN Y J , HAN S , MAO H Z ,et al. Deep gradient compression:reducing the communication bandwidth for distributed training[J]. arXiv preprint,2017,arXiv:1712.01887.
[17]	DAI W , KUMAR A , WEI J ,et al. Highperformance distributed ML at scale through parameter server consistency models[C]// AAAI Conference on Artificial Intelligence. New York:ACM Press, 2015.
[18]	RECHT B , RE C , WRIGHT S ,et al. Hogwild:a lock-free approach to parallelizing stochastic gradient descent[C]// Advances in Neural Information Processing Systems.[S.l.:s.n]. 2011: 693-701.
[19]	HO Q , CIPAR J , CUI H G ,et al. More effective distributed ml via a stale synchronous parallel parameter server[C]// Advances in Neural Information Processing Systems.[S.l.:s.n]. 2013: 1223-1231.
[20]	FENG S W , YU H . Multi-participant multi-class vertical federated learning[J]. arXiv preprint,2020,arXiv:2001.11154.
[21]	KONE?NY J . Stochastic distributed and federated optimization for machine learning[J]. arXiv preprint,2017,arXiv:1707.01155.
[22]	LIU X Y , LI H W , XU G W ,et al. Adaptive privacy-preserving federated learning[J]. Peer-to-Peer Networking and Applications. 2020
[23]	HU R , GONG Y M , GUO Y X . CPFed:communication-efficient and privacypreserving federated learning[J]. arXiv preprint,2020,arXiv:2003.13761.
[24]	RYFFEL T , TRASK A , DAHL M ,et al. A generic framework for privacy preserving deep learning[J]. arXiv preprint,2018,arXiv:1811.04017.
[25]	ANTONIOUS M , DEEPESH D , SUHAS D ,et al. Shuffled model of federated learning:privacy,communication and accuracy trade-offs[J]. arXiv preprint,2020,arXiv:008.07180.
[26]	SMITH V , CHIANG C K , SANJABI M ,et al. Federated multi-task learning[C]// Advances in Neural Information Processing Systems.[S.l.:s.n]. 2017: 4424-4434.
[27]	CORINZIA L , BUHMANN J M . Variational federated multi-task learning[J]. arXiv preprint,2019,arXiv:1906.06268.
[28]	CALDAS S , SMITH V , TALWALKAR A . Federated kernelized multi-task learning[C]// SysML Conference 2018.[S.l.:s.n]. 2018.
[29]	KALLMAN R , KIMURA H , NATKINS J ,et al. H-store:a high-performance,distributed main memory transaction processing system[J]. Proceedings of the VLDB Endowment, 2008,1(2): 1496-1499.
[30]	YANG K , JIANG T , SHI Y M ,et al. Federated learning via over-the-air computation[J]. IEEE Transactions on Wireless Communications, 2020,19(3): 2022-2035.
[31]	NISHIO T , YONETANI R . Client selection for federated learning with heterogeneous resources in mobile edge[C]// 2019 IEEE International Conference on Communications. Piscataway:IEEE Press, 2019: 1-7.
[32]	WANG J Y , SAHU A K , YANG Z Y ,et al. MATCHA:speeding up decentralized SGD via matching decomposition sampling[J]. arXiv preprint,2019,arXiv:1905.09435.
[33]	REISIZADEH A , MOKHTARI A , HASSANI H ,et al. Fedpaq:a communication-efficient federated learning method with periodic averaging and quantization[J]. arXiv preprint,2019,arXiv:1909.13014.
[34]	KHALED A , MISHCHENKO K , RICHTáRIK P . Better communication complexity for local SGD[J]. arXiv preprint,2019,arXiv:1909.04746.
[35]	LI S Y , CHENG Y , LIU Y ,et al. Abnormal client behavior detection in federated learning[J]. arXiv preprint,2019,arXiv:1910.09933.
[36]	SATTLER F , WIEDEMANN S,MüLLER K R ,et al. Robust and communicationefficient federated learning from nonIID data[J]. IEEE Transactions on Neural Networks and Learning Systems. 2019
[37]	C ROTTY A , GALAKATOS A , KRASKA T . Tupleware:distributed machine learning on small clusters[J]. IEEE Data Engineering Bulletin, 2014,37(3): 63-76.
[38]	JOLFAEI A , OSTOVARI P , ALAZAB M ,et al. Guest editorial special issue on privacy and security in distributed edge computing and evolving IoT[J]. IEEE Internet of Things Journal, 2020,7(4): 2496-2500.
[39]	SAHU A K , LI T , SANJABI M ,et al. Federated optimization for heterogeneous networks[J]. arXiv preprint,2018,arXiv:1812.06127.
[40]	YANG K , FAN T , CHEN T J ,et al. A quasi-newton method based vertical federated learning framework for logistic regression[J]. arXiv preprint,2019,arXiv:1912.00513.
[41]	YANG S W , REN B , ZHOU X H ,et al. Parallel distributed logistic regression for vertical federated learning without thirdparty coordinator[J]. arXiv preprint,2019,arXiv:1911.09824.
[42]	GAO D S , JU C , WEI X G ,et al. HHHFL:hierarchical heterogeneous horizontal federated learning for electroencephalography[J]. arXiv preprint,2019,arXiv:1909.05784.
[43]	LIU Y , KANG Y , ZHANG X W ,et al. A communication efficient vertical federated learning framework[J]. arXiv preprint,2019,arXiv:1912.11187.
[44]	SHARMA S , XING C P , LIU Y ,et al. Secure and efficient federated transfer learning[J]. arXiv preprint,2019,arXiv:1910.13271.
[45]	ZHAO Y , LI M , LAI L Z ,et al. Federated learning with non-IID data[J]. arXiv preprint,2018,arXiv:1806.00582.
[46]	LIU Y , LIU Y T , LIU Z J ,et al. Federated forest[J]. IEEE Transactions on Big Data,2020:Accepted.
[47]	CHENG K W , FAN T , JIN Y L ,et al. SecureBoost:a lossless federated learning framework[J]. arXiv preprint,2019,arXiv:1901.08755.
[48]	LI Q B , WEN Z Y , HE B S . Practical federated gradient boosting decision trees[J]. arXiv preprint,2019,arXiv:1911.04206.
[49]	HARTMANN V , MODI K , PUJOL J M ,et al. Privacy-preserving classification with secret vector machines[J]. arXiv preprint,2019,arXiv:1907.03373.
[50]	ZHU X H , WANG J , HONG Z ,et al. Federated learning of unsegmented Chinese text recognition model[C]// 2019 IEEE 31st International Conference on Tools with Artificial Intelligence. Piscataway:IEEE Press, 2019: 1341-1345.
[51]	BHOWMICK A , DUCHI J , FREUDIGER J ,et al. Protection against reconstruction and its applications in private federated learning[J]. arXiv preprint,2018,arXiv:1812.00984.
[52]	DUCHI J , JORDAN M I , MCMAHAN B . Estimation,optimization,and parallelism when data is sparse[C]// In Advances in Neural Information Processing Systems. New York:ACM Press, 2013.
[53]	CHILIMBI T , SUZUE Y , APACIBLE J ,et al. Project adam:building an efficient and scalable deep learning training system[C]// The 11th USENIX Symposium on Operating Systems Design and Implementation. New York:ACM Press, 2014: 571-582.
[54]	LIU Y , MUPPALA J K , VEERARAGHAVAN M ,et al. Data center networks:topologies,architectures and fault-tolerance characteristics[M]. Heidelberg: Springer Science ＆ Business MediaPress, 2013.
[55]	BONAWITZ K , EICHNER H , GRIESKAMP W ,et al. Towards federated learning at scale:system design[J]. arXiv preprint,2019,arXiv:1902.01046.
[56]	LI X , HUANG K , YANG W ,et al. On the convergence of FedAvg on non-IID data[J]. arXiv preprint,2019,arXiv:1907.02189.
[57]	CALDAS S,KONE?NY J , MCMAHAN H B ,et al. Expanding the reach of federated learning by reducing client resource requirements[J]. arXiv preprint,2018,arXiv:1812.07210.
[58]	NISHIO T , YONETANI R . Client selection for federated learning with heterogeneous resources in mobile edge[C]// ICC 20192019 IEEE International Conference on Communications. Piscataway:IEEE Press, 2019: 1-7.
[59]	YOSHIDA N , NISHIO T , MORIKURA M ,et al. Hybrid-FL for wireless networks:cooperative learning mechanism using non-IID data[J]. arXiv preprint,2019,arXiv:1905.07210.
[60]	SPRAGUE M R , JALALIRAD A , SCAVUZZO M ,et al. Asynchronous federated learning for geospatial applications[C]// Joint European Conference on Machine Learning and Knowledge Discovery in Databases. Heidelberg:Springer, 2018: 21-28.
[61]	XIE C , KOYEJO S , GUPTA I . Asynchronous federated optimization[J]. arXiv preprint,2019,arXiv:1903.03934.
[62]	YANG J L , DUAN Y X , QIAO T ,et al. Prototyping federated learning on edge computing systems[J]. Frontiers of Computer Science, 2020,14: 1-3.
[63]	WANG S Q , TUOR T , SALONIDIS T ,et al. Adaptive federated learning in resource constrained edge computing systems[J]. IEEE Journal on Selected Areas in Communications, 2019,37(6): 1205-1221.
[64]	ZHAO Y , ZHAO J , JIANG L S ,et al. Mobile edge computing,blockchain and reputation-based crowd-sourcing IoT federated learning:a secure,decentralized and privacy-preserving system[J]. arXiv preprint,2019,arXiv:1906.10893.
[65]	LI Z Y , LIU J , HAO J L ,et al. CrowdSFL:a secure crowd computing framework based on blockchain and federated learning[J]. Electronics, 2020,9(5):773.
[66]	KANG J W , XIONG Z H , NIYATO D ,et al. Incentive design for efficient federated learning in mobile networks:a contract theory approach[C]// 2019 IEEE VTS Asia Pacific Wireless Communications Symposium. Piscataway:IEEE Press, 2019: 1-5.
[67]	ISAKSSON M , NORRMAN K . Secure federated learning in 5G mobile networks[J]. arXiv preprint,2020,arXiv:2004.06700.

类型	基础	算法	框架	特点
联邦机器学习	联邦线性算法	逻辑回归^[40]	中心	同态加密，观察模型变化，周期性梯度更新
		逻辑回归^[41]	去中心	取消第三方参与，有标签数据持有方主导，差分隐私
	联邦树模型	联邦森林^[46]	中心	模型分散存储，中心服务器储存结构
		梯度上升树SecureBoost^[47]	去中心	同态加密，特征分桶聚合，保障准确率
		梯度上升树SimFL^[48]	去中心	哈希表加密，加权梯度上升，通信效率高
	联邦支持向量机	支持向量机Valentin^[49]	中心	哈希表加密，次梯度更新，隐私性较好
联邦深度学习	联邦神经网络	NN^[8]	中心	比传统神经网络收敛更快，参数联合初始化时具有更好的收敛效果
	联邦卷积神经网络	CNN^[55]	中心	网络结构比RNN简单，收敛速度更快
		VGG11^[51]	中心	non-IID数据上，参数压缩的优化算法收敛效果较差；不压缩的收敛效果较好，但参数量较大
	联邦LSTM	LSTM^[8,39,51]	中心	受数据分布影响较大，不同的参数聚合方式效果不同

优化角度	文献方法	优化方法	优缺点
通信成本	FedAvg^[8]	IID 数据；增加参与方本地计算	增加计算成本；non-IID数据优化效果差
	FedProx^[39]	non-IID数据；增加本地计算	增加计算成本，可优化non-IID数据，代价是准确性降低
	VFL^[54]	纵向联邦算法；增加本地计算	增加计算成本，代价是降低准确性
	结构和轮廓更新机制^[9]	压缩传输模型，提升参与方到服务器的通信效率	参与方到服务器参数压缩，代价是复杂的模型结构可能出现收敛问题
	服务器-客户端更新^[57]	压缩传输模型，提升服务器到参与方的通信效率	服务器到参与方参数压缩，代价是准确性降低，可能有收敛问题
客户端选择	FedCS^[58]	选择迭代效率最优的模型训练参与方	比FedAvg更准确，但是只能被应用于简单的NN模型，不适合复杂模型
	Hybrid-FL^[59]	服务器选择客户端数据组成近似IID的数据集	non-IID数据收敛有问题
异步聚合	AsyncFedAvg^[60]	服务器接收到客户端参数更新就立刻聚合	存在non-IID数据收敛问题
	FedAsync^[61]	服务端通过加权聚合的方式获取客户端的模型参数	难调参数，存在收敛问题

联邦学习算法综述

Research review of federated learning algorithms

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 4

参考文献 67

相关文章 15

Metrics

推荐阅读 0

[1]	钱海红, 王茂异, 熊贇. 高等教育数字化转型的现状与发展研究[J]. 大数据, 2023, 9(3): 56-70.
[2]	张传尧, 司世景, 王健宗, 肖京. 联邦元学习综述[J]. 大数据, 2023, 9(2): 122-146.
[3]	梅宏, 杜小勇, 金海, 程学旗, 柴云鹏, 石宣化, 靳小龙, 王亚沙, 刘驰. 大数据技术前瞻[J]. 大数据, 2023, 9(1): 1-20.
[4]	沈阳, 余梦珑. 元宇宙与大数据：时空智能中的数据洞察与价值连接[J]. 大数据, 2023, 9(1): 103-110.
[5]	陈静. 人文大数据及其在数字人文领域中的应用[J]. 大数据, 2022, 8(6): 3-14.
[6]	罗煜楚, 吴昊, 郭宇涵, 谭绍聪, 刘灿, 蒋瑞珂, 袁晓如. 数字人文中的可视化[J]. 大数据, 2022, 8(6): 74-93.
[7]	郑童哲恒, 李斌, 冯敏萱, 常博林, 王东波. 历史典籍的结构化探索——《史记·列传》数字人文知识库的构建与可视化研究[J]. 大数据, 2022, 8(6): 40-55.
[8]	张燕, 杨一帆, 伊人, 罗圣美, 唐剑飞, 夏正勋. 隐私计算场景下数据质量治理探索与实践[J]. 大数据, 2022, 8(5): 55-73.
[9]	尹虹舒, 周旭华, 周文君. 纵向联邦线性模型在线推理过程中成员推断攻击的隐私保护研究[J]. 大数据, 2022, 8(5): 45-54.
[10]	吴建汉, 司世景, 王健宗, 肖京. 联邦学习攻击与防御综述[J]. 大数据, 2022, 8(5): 12-32.
[11]	朱智韬, 司世景, 王健宗, 肖京. 联邦推荐系统综述[J]. 大数据, 2022, 8(4): 105-132.
[12]	李汶龙, 袁媛, 安筱鹏. 刍议大数据治理的三大基础思维[J]. 大数据, 2022, 8(4): 34-45.
[13]	汤奇峰, 邵志清, 叶雅珍. 数据交易中的权利确认和授予体系[J]. 大数据, 2022, 8(3): 40-53.
[14]	王陈慧子, 蔡玮. 元宇宙数字经济：现状、特征与发展建议[J]. 大数据, 2022, 8(3): 140-150.
[15]	杨玫, 李玮, 乔思渊, 刘巍. 中国大数据产业产值测算方法研究[J]. 大数据, 2022, 8(3): 151-160.