期刊名称：IAENG International Journal of Computer Science
印刷版ISSN：1819-656X
电子版ISSN：1819-9224
出版年度：2019
卷号：46
期号：4
页码：512-517
出版社：IAENG - International Association of Engineers
摘要：Cloud (computing) system has been widely used invarious fields. However, as the number of terminals increases,the limits of the capabilities are also becoming clear. Thelimits lead to the delay of significant processing time. Inorder to improve this, the edge (or fog) computing systemhas been proposed. In the conventional cloud system, users(clients) send all data to the cloud and the cloud returns thecomputation result to users (clients). On the other hand, inthe edge (computing) system, multiple servers called edges areassigned between the cloud and the terminals (or things). In thesystem, there are two types of servers for cloud and edge. Heavyand normal tasks are processed in the cloud and edge servers,respectively. Then, how is machine learning in the cloud or edgesystem? The purpose of learning is to find out the relationship(information) lurking in from the collected data. That is, asystem with several parameters is assumed and estimated byrepeatedly updating the parameters with learning data. Further,there is the problem of security for learning data. How can webuild the cloud system to avoid such risk? Secure multipartycomputation (SMC) is known as one method realizing securecomputation. Many studies on learning methods based on SMChave also been proposed in the cloud system. Then, what kindof learning method is suitable for the edge system based onSMC? In this paper, Neural Gas (NG) algorithms to realizefast and secure processing on edge computing for clusteringand classification problems are proposed and the effectivenessof the proposed methods in numerical simulations is shown.
关键词：IoT; Machine learning; Security; Batch learning;Clustering; Classification problem; Neural Gas
