摘要
基于非线性降维技术有助于发现高维数据内在结构与几何分布的理论基础,根据特征保留形式将目前非线性降维技术分为3类,并对其中具有代表性的算法进行分析。通过与经典线性降维技术进行对比,证明了非线性降维技术在数据可视化应用中的优势。针对传统非线性降维技术存在的时间复杂度过高及适用范围有限的问题,系统性地总结了目前该领域的最新改进方式。
Based on the theory that nonlinear dimension reduction technology is a helpful method to discover the intrinsic structure and geometric distribution of high-dimensional data,the typical nonlinear dimensionality reduction techniques were divided into three kinds according to the forms of feature retention and they were analyzed respectively.Compared with the classical linear dimension reduction technology,it was proved that the nonlinear dimension reduction technology has good performance in the application of data visualization.Aiming at high time complexity and incomplete applicability of traditional nonlinear dimension reduction technology,the latest improvements in this field were systematically summarized.
作者
杜杰
王骁
胡良剑
DU Jie;WANG Xiao;HU Liangjian(College of Science,Donghua University,Shanghai 201620,China)
出处
《东华大学学报（自然科学版）》
CAS
北大核心
2020年第4期675-680,共6页
Journal of Donghua University(Natural Science)
基金
国家级大学生创新创业训练资助项目(201810255012)。
关键词
非线性
特征保留
高维数据
降维技术
可视化
nonlinear
feature retention
high-dimensional data
dimension reduction technology
visualization
