首页 | 本学科首页   官方微博 | 高级检索  
     

基于流形弯曲度的有序自适应邻域选择算法
引用本文:李德玉,高翠珍,翟岩慧. 基于流形弯曲度的有序自适应邻域选择算法[J]. 山西大学学报(自然科学版), 2012, 35(2): 219-223
作者姓名:李德玉  高翠珍  翟岩慧
作者单位:1. 山西大学计算机与信息技术学院,山西太原030006;山西大学计算智能与中文信息处理教育部重点实验室,山西太原030006
2. 山西大学计算机与信息技术学院,山西太原,030006
基金项目:国家自然科学基金,教育部高等学校博士点基金,山西省自然科学基金,山西省科技攻关项目
摘    要:针对传统邻域选择方法不能根据流形样本密度和弯曲度合理选择邻域的缺点,提出了一种有序自适应的邻域选择算法.该算法从流形上曲率最小的点开始,以宽度优先的次序不断地处理每个点.对搜索到的数据点,基于流形结构的局部线性特性,利用已有的邻域信息估算其局部切空间,然后通过其邻域边在切空间的投影自适应地选择合适的邻域.实验结果表明:该算法应用于Isomap后,对不同结构的数据集嵌入结果更准确.

关 键 词:流形学习  邻域选择  切空间

An Orderly Adaptive Neighborhood Selection Algorithm Based on Manifold Curvature
LI De-yu , GAO Cui-zhen , ZHAI Yan-hui. An Orderly Adaptive Neighborhood Selection Algorithm Based on Manifold Curvature[J]. Journal of Shanxi University (Natural Science Edit, 2012, 35(2): 219-223
Authors:LI De-yu    GAO Cui-zhen    ZHAI Yan-hui
Affiliation:1(1.School of Computer & Information Technology,Shanxi University,Taiyuan 030006,China; 2.Key Laboratory of Computational Intelligence and Chinese Information Processing of Ministry of Education,Shanxi University,Taiyuan 030006,China)
Abstract:An orderly adaptive neighborhood selection algorithm is introduced because the traditional neighborhood selection algorithm has a defect that can not select neighbors adaptively based on the sample density and curvature.In this algorithm,data points start with the smallest curvature point in the manifold,and breadth-first searching algorithm was used to expand manifold data.For each point,we estimate the local tangent space based on the local linearity of manifold structure with existing neighborhood and then choose the right neighborhood adaptively through mapping of the neighborhood edge in the tangent space.This method is applied to Isomap,and experimental results validate the accurate of the embedding results for different data sets.
Keywords:manifold learning  neighborhood selection  tangent space
本文献已被 CNKI 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号