[1]徐胜超.流形学习降维算法中一种新动态邻域选择方法[J].计算机技术与发展,2022,32(01):85-90.[doi:10. 3969 / j. issn. 1673-629X. 2022. 01. 015]
　XU Sheng-chao.A Dynamic Neighborhood Selection Approach for ManifoldLearning Dimensionality Reduction Algorithm[J].,2022,32(01):85-90.[doi:10. 3969 / j. issn. 1673-629X. 2022. 01. 015]
点击复制

流形学习降维算法中一种新动态邻域选择方法()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
32
期数:
2022年01期
页码:
85-90
栏目:
大数据分析与挖掘
出版日期:
2022-01-10

文章信息/Info

Title:
A Dynamic Neighborhood Selection Approach for ManifoldLearning Dimensionality Reduction Algorithm
文章编号:
1673-629X(2022)01-0085-06
作者:
徐胜超
广州华商学院 数据科学学院,广东 广州 511300
Author(s):
XU Sheng-chao
School of Date Science,Guangzhou Huashang College,Guangzhou 511300,China
关键词:
流形学习; 黑塞局部线性嵌入; 数据挖掘; 降维算法; 相邻区域
Keywords:
manifold learning; Hessian locally linear embedding ( LLE) ; data mining; dimensionality reduction algorithm; neighborhood
分类号:
TP391. 1
DOI:
10. 3969 / j. issn. 1673-629X. 2022. 01. 015
摘要:
近年来,高维数据算法在诸如机器学习领域以及模式识别当中有着十分广泛的应用。 降维算法的目的是为了揭示出在高维数据空间中样本数据的固有的组成特性,关注于寻找原始数据集特征表示中有价值的信息。 相邻区域选择问题对流形学习降维算法的性能改进至关重要。 因此,该文提出一种流形学习降维算法中的新动态邻域选择方法 Mod -HLLE( modified Hessian locally linear embedding) 。 该方法针对 Hessian 布局线嵌入方法 HLLE 进行了考察,Mod-HLLE 算法是针对高维数据的局部线性嵌入降维算法的改进。 Mod-HLLE 主要通过计算每个数据点的局部相邻区域参数的方式来完成测量距离和欧几里德距离的评测,再通过动态的相邻区域的尺寸大小来选择新的局部相邻区域。 Mod-HLLE 在非噪声干扰和噪声干扰情况下,对两类典型 3D 高维数据集进行降维测试。 实验结果表明,Mod-HLLE 可以获得很好的几何直观效果,在性能和稳定性方面都优于常见的降维算法,对其他高维数据降维算法的改进也具有很好的参考价值。
Abstract:
In recent years,high-dimensional data algorithms have been widely used in machine learning and pattern recognition. The aimof dimensionality reduction is to reveal the intrinsic composition of the distribution of samples in the initiate high-dimensional space andfind the valuable information in the original data set feature representation. The neighborhood selection is quite important to improve theperformance of manifold learning dimensionality reduction algorithm. Therefore,we propose a dynamic neighborhood selection approachfor manifold learning dimensionality reduction algorithm call Mod-HLLE ( modified Hessian locally linear embedding) . It is used to investigate the Hessian layout line embedding method ( HLLE) . Mod-HLLE is an improvement of local linear embedding dimensionalityreduction algorithm for high dimensional data. Mod - HLLE mainly evaluates the measurement distance and Euclidean distance bycalculating the local adjacent area parameters of each data point,and then selects the new local adjacent area by the size of the dynamicadjacent area. Two kinds of 3D dimensionality data sets which are often unevenly distributed are adopted in noise case and without noisecase. Experiment shows that Mod-HLLE can obtain ideal geometric intuitive effect,which is superior to common dimensional reductionalgorithms in terms of performance and stability, and also has a good reference value for improving other high - dimensional datadimensional reduction algorithms.

相似文献/References:

[1]闫志敏 刘希玉.流形学习及其算法研究[J].计算机技术与发展,2011,(05):99.
　YAN Zhi-min,LIU Xi-yu.Manifold Learning and Research of Algorithm[J].,2011,(01):99.
[2]王健,韩志艳.鲁棒主元分析及在故障诊断中的应用[J].计算机技术与发展,2013,(06):79.
　WANG Jian[],HAN Zhi-yan[].Robust Principal Component Analysis and Its Application in Fault Diagnosis[J].,2013,(01):79.
[3]吴飞,荆晓远,李文倩,等.基于流形学习的整体正交稀疏保留鉴别分析[J].计算机技术与发展,2014,24(06):63.
　WU Fei,JING Xiao-yuan,LI Wen-qian,et al.Analysis of Preserving Discriminant of Holistic Orthogonal Sparsity Based on Manifold Learning[J].,2014,24(01):63.
[4]刘茜[],荆晓远[],李文倩[],等. 基于流形学习的正交稀疏保留投影[J].计算机技术与发展,2014,24(07):34.
　LIU Qian[],JING Xiao-yuan[,LI Wen-qian[],et al. Orthogonal Sparsity Preserving Projections Based on Manifold Learning[J].,2014,24(01):34.
[5]凌若冰[],荆晓远[],吴飞[],等. 基于流形学习的正交稀疏保留投影鉴别分析[J].计算机技术与发展,2015,25(01):66.
　LING Ruo-bing[],JING Xiao-yuan[],WU Fei[],et al. Orthogonal Sparsity Preserving Discriminant Analysis Based on Manifold Learning[J].,2015,25(01):66.
[6]刘丹霞,干宗良,杨文峰. 基于相似性约束的人脸超分辨率重建算法[J].计算机技术与发展,2015,25(08):58.
　LIU Dan-xia,GAN Zong-liang,YANG Wen-feng. A Face Super-resolution Reconstruction Algorithm Based on Similarity Constraints[J].,2015,25(01):58.
[7]杨秋颖,翁小清.基于 LLE 和高斯混合模型的时间序列聚类[J].计算机技术与发展,2022,32(08):33.[doi:10. 3969 / j. issn. 1673-629X. 2022. 08. 006]
　YANG Qiu-ying,WENG Xiao-qing.Time Series Clustering Based on LLE and Gaussian Mixture Model[J].,2022,32(01):33.[doi:10. 3969 / j. issn. 1673-629X. 2022. 08. 006]
[8]李燕燕,闫德勤.一种优化的近邻保持嵌入降维算法研究[J].计算机技术与发展,2023,33(06):28.[doi:10. 3969 / j. issn. 1673-629X. 2023. 06. 005]
　LI Yan-yan,YAN De-qin.Research on an Optimized Nearest Neighbor Preserving Embedding Algorithm for Dimensionality Reduction[J].,2023,33(01):28.[doi:10. 3969 / j. issn. 1673-629X. 2023. 06. 005]

更新日期/Last Update: 2022-01-10