复杂仿真增量数据可视化中的降维研究  

Research on Dimension Reduction for Complex Simulation Incremental Data Visualization

作　　者：李惠君[1,2] 张利辉[1,3] 刘雪飞[1] 唐勇[3] 

机构地区：[1]中国环境管理干部学院信息工程系,秦皇岛066012 [2]燕山大学电气工程学院,秦皇岛066004 [3]燕山大学信息工程学院,秦皇岛066004

出　　处：《系统仿真学报》2013年第10期2278-2284,共7页Journal of System Simulation

基　　金：国家自然科学基金(60970073);河北省自然基金资助项目(F2012203084)

摘　　要：降维是复杂仿真数据可视化中一个重要环节,它解决了特征提取问题。但由于复杂仿真数据结构复杂、往往呈非线性特点,无法采用传统的线性方法进行降维。提出了一种全新的基于流形学习的复杂仿真数据降维方法,并在对常用流形学习降维方法分析的基础上,给出了局部切空间排列算法的基于核方法的改进,使之能够进行仿真增量数据可视化中的降维。最终经试验验证该方法切实有效。Dimension reduction is an important step in the complex simulation data visualization. It solves the problem of feature extraction, But due to the complex structure and the nonlinear characteristics of the complex simulation data, the traditional linear methods often can＇t be used for dimension reduction. A new dimension reduction method based on manifold learning for complex simulation data visualization was proposed. By commonly used manifold learning dimension reduction methods, the local tangent space alignment algorithm based on kernel method was improved. The dimensions of the complex simulation incremental data were reduced. The final experiments have verified the validity of the new method.

关 键 词：流形学习 降维 增量数据 复杂仿真 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]