基于粒计算的大数据处理  

Review of Big Data Processing Based on Granular Computing

作　　者：徐计[1,2,3] 王国胤[3] 于洪[2] 

机构地区：[1]西南交通大学信息科学与技术学院,成都610031 [2]重庆邮电大学计算智能重庆市重点实验室,重庆400065 [3]中国科学院重庆绿色智能技术研究院,电子信息技术研究所,重庆401122

出　　处：《计算机学报》2015年第8期1497-1517,共21页Chinese Journal of Computers

基　　金：国家自然科学基金(61272060,61379114);重庆市自然科学基金重点项目(CSTC2013jjB40003)资助~~

摘　　要：在大数据时代,如何充分挖掘出蕴藏于数据资源中的价值正在成为各国IT业界、学术界和政府共同关注的焦点.使用云计算平台分布式地存储和分析大数据已经成为共识并且得到了广泛应用,但这并没有完全解决大数据的3V特性带来的问题.全面应对大数据的挑战需要来自存储技术、下一代网络、处理器、计算模型等各个领域的创新.粒计算是在求解问题过程中使用"粒"的理论、方法、技术和工具的集合,适用于近似求解有不确定性和层次结构的问题.该文综述了大数据处理的研究现状,分析了当前大数据处理研究存在的局限性;根据运用粒计算方法解决问题的不同特征,该文归纳了粒计算的3种基本模式,回顾了各种模式的相关研究工作;该文讨论了粒计算应用于大数据处理的可行性与优势,并探讨了在大数据的粒计算处理框架中需要解决的各个关键问题.In big data age,people from IT industry,academia and government are taking effort to find out the value behind the data resources.Using cloud computing platform to store and process big data distributed has been a common sense and been put into practice extensively.However,this is far from solving the problems completely caused by the 3V-features（volume,variety and velocity）of big data.The problems require joint efforts of researchers from the fields of storage technology,Internet2,processors,and computing paradigm,etc.Granular computing（GrC）is a label of the family of any theories,methodologies,techniques,and tools that make use of information granules in problem solving,especially when the problem is with uncertainty,of hierarchical structure and can be approximately solved.This paper surveys the researches on big data processing and analyzes the limitation in current big data processing researches,summarizes three basic modes of GrC according to the ways in which GrC is used to solve problems and reviews the related works of each mode.The feasibility and advantages of applying GrC to process big data are discussed subsequently,and the related key problems rising in GrC-based big data processing framework are discussed at last.

关 键 词：粒计算 大数据 云计算 深度学习 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]