三元空间大数据网络关联表征

Cyber-physical-human big data correlational representation

三元空间是由信息空间、物理世界、人类社会所构成,三元空间大数据由信息空间所产生的互联网数据、物理空间所产生的物联网数据和人类社会空间所产生的数据所构成.本文介绍了三元空间大数据的关联复杂性,并针对三元空间大数据关联复杂性这一本质困难,提出解决三元空间异构数据的关联表征这一关键科学问题.传统信息科学往往依据既有的先验信息进行特征表达,并在先验表征空间内分析信息语义及其关联关系.然而,在表征层面所引入的先验偏见,极大限制了信息理解和关联的广度和深度,使得蕴含于三元空间大数据之中的超出人类现有经验的知识难以被发掘.为解决上述难题,本文提出将三元空间数据关联关系通过拓扑图理论表征成网络或图以实现三元空间大数据的关联表征和融合分析.具体地,利用数据驱动的深层网络表征对三元空间异构数据的弱先验关联关系进行深度建模以保持原始数据间的异构关联关系,通过对非线性、非凸异构数据关联表达进行深度解离化计算以提升深层关联表征的鲁棒性与可解释性,借助拓扑图理论挖掘三元空间大数据中蕴含的异构关联关系以达到对三元空间异构关系的精准刻画.最后,本文从知识与数据双驱动、自适应,以及可推理三元空间大数据分析的角度对三元空间大数据关联表征的未来研究方向进行展望.

Cyber-physical-human ternary space is composed of information space,physical world and human society.Cyber-physical-human ternary space big data consists of the Internet data generated by the information space,the IoT data generated in physical space and data generated in human social space.This paper introduces the associated complexity of ternary spatial big data.Moreover,in view of the inherent difficulties lying in the correlation complexity of big data in ternary space,we propose to solve the key scientific problem of correlational representation for heterogeneous ternary space data.We take full advantages of deep representations obtained by deep architectures to model the heterogeneous correlational relationship of ternary space data,and establish a joint data-knowledge dual driven ternary space big data analysis theory.Last but not least,we present our insights on future research directions that deserve further investigation in the area of cyber-physical-human ternary space big data analysis.