Federated Learning with Class Imbalance Reduction

doi:10.23919/EUSIPCO54536.2021.9616052

	Federated Learning with Class Imbalance Reduction
	Yang, Miao1,2 ; Wang, Ximin1 ; Zhu, Hongbin1 ; Wang, Haifeng 3; Qian, Hua1,2
	2021-08-23
会议录名称	2021 29TH EUROPEAN SIGNAL PROCESSING CONFERENCE (EUSIPCO)
ISSN	2076-1465
发表状态	已发表
DOI	10.23919/EUSIPCO54536.2021.9616052
摘要	Federated learning (FL) is a promising technique that enables a large amount of edge computing devices to collaboratively train a global learning model. Due to the communication limitation, only a subset of devices can be engaged to train and transmit the trained model to centralized server for aggregation. Since the local data distribution varies among all devices, class imbalance problem arises along with the unfavorable client selection, resulting in a slow converge rate of the global model. In this paper, we design an estimation scheme to reveal the class distribution without the awareness of raw data. According to the estimation scheme, we propose a multi-arm bandit based algorithm that can select the client set with minimal class imbalance. The proposed algorithm can significantly improve the convergence performance of the global model. Simulation results demonstrate the effectiveness of the proposed algorithm.
关键词	federated learning deep neural networks privacy concerns class imbalance client scheduling multi-armed bandit
会议名称	29th European Signal Processing Conference (EUSIPCO)
出版地	PO BOX 74251, KESSARIANI, 151 10, GREECE
会议地点	null,null,ELECTR NETWORK
会议日期	AUG 23-27, 2021
URL	查看原文
收录类别	EI ; CPCI ; CPCI-S
语种	英语
资助项目	National Natural Science Foundation of China[61671436] ; Science and Technology Commission Foundation of Shanghai[19DZ1204300]
WOS研究方向	Acoustics ; Computer Science ; Engineering ; Imaging Science & Photographic Technology ; Telecommunications
WOS类目	Acoustics ; Computer Science, Software Engineering ; Engineering, Electrical & Electronic ; Imaging Science & Photographic Technology ; Telecommunications
WOS记录号	WOS:000764066600430
出版者	EUROPEAN ASSOC SIGNAL SPEECH & IMAGE PROCESSING-EURASIP
引用统计	正在获取...
文献类型	会议论文
条目标识符	https://kms.shanghaitech.edu.cn/handle/2MSLDSTB/135532
专题	信息科学与技术学院信息科学与技术学院_特聘教授组_钱骅组信息科学与技术学院_硕士生信息科学与技术学院_博士生
通讯作者	Yang, Miao; Qian, Hua
作者单位	1.ShanghaiTech Univ, Sch Informat Sci & Technol, Shanghai, Peoples R China 2.Chinese Acad Sci, Shanghai Adv Res Inst, Shanghai, Peoples R China 3.Chinese Acad Sci, Key Lab Wireless Sensor Network Commun, Beijing, Peoples R China
第一作者单位	信息科学与技术学院
通讯作者单位	信息科学与技术学院
第一作者的第一单位	信息科学与技术学院
推荐引用方式 GB/T 7714	Yang, Miao,Wang, Ximin,Zhu, Hongbin,et al. Federated Learning with Class Imbalance Reduction[C]. PO BOX 74251, KESSARIANI, 151 10, GREECE:EUROPEAN ASSOC SIGNAL SPEECH & IMAGE PROCESSING-EURASIP,2021.