摘要
随着信息技术的飞速发展,现实世界中涌现出大量的跨媒体数据.所谓跨媒体数据是指那些表达的内容相似,但以不同模态、不同来源、不同背景等形式出现的数据.比如,一张描述花豹的网页通常采用共生的图片和文本等不同的模态刻画花豹的外形和习性.这些跨媒体数据通常呈现出底层特征异构、高层语义相关的特性.传统的单媒体学习方法已无法适应跨媒体数据呈现出的特征异构性.因此,跨媒体学习相关理论与方法的研究是当前数字媒体分析领域的热点研究课题之一.该文主要介绍了跨媒体学习的研究背景和应用价值,概括介绍了各类跨媒体学习相关方法的数学原理和基本特性,并重点介绍了跨媒体共享子空间学习的研究进展,比较了基于投影、矩阵分解、任务和度量等四大类子空间学习方法的优缺点,分析了未来的发展方向.
With the rapid development of information technology, there are many CMD （Cross- Media Data） in the real world. The so-called cross-media data refer to information items with similar underlying contents, which arrive in different modalities, sources or backgrounds, and so on. For example, a webpage describing leopards uses co-occurring text and image of different modalities to represent leopards. These CMD show characteristics of the heterogeneity of low- level features and the correlation of high-level semantics. Traditional Mono-Media Learning （MML） methods have not been able to adapt to the feature heterogeneity of CMD. Thus, the research on related problems in CML （Cross-Media Learning） have been One of the hot research topics in the field of digital media analysis recently. This paper mainly introduces the research background and application value of CML, and provides an overview of mathematical principle and specialties of various related methods in CML. Meanwhile, the advances in cross-media shared subspace learning are presented. Furthermore, the paper compares the advantages and disadvantages of four kinds of subspace learning task and measurement, respectively. Finally. methods based on projection, matrix decomposition, the future development of CMI. is analyzed.
作者
张磊
赵耀
朱振峰
ZHANG Lei ZHAO Yao ZHU Zhen-Feng(institute of Information Engineering, Chinese Academy of Sciences, Beijing 100093 institute of Information Science, Beijing Jiaotong University, Beijing 100044 Beijing Key Laboratory of Advanced Information Science and Network Technology, Beijing Jiaotong University, Beijing 100044)
出处
《计算机学报》
EI
CSCD
北大核心
2017年第6期1394-1421,共28页
Chinese Journal of Computers
基金
国家"九七三"重点基础研究发展规划项目基金(2012CB316401)
国家自然科学基金(61532005
61572068
61271275
61501457
61601458
61602465)
长江学者和创新研究团队项目(IRT201206)
新世纪优秀人才支持计划(13-0661)资助~~
关键词
跨媒体
异构数据
共享子空间
多视角学习
优化
人工智能
cross-media
heterogeneous data
shared subspace
multi-view learning
optimization： arlificiall inlelligence
