资源分类:Matlab 工具:MATLAB 7.13 (R2011b)

SpectralClustering实现了三个谱聚类算法(Unnormalized, Shi & Malik, Jordan & Weiss).

Spectral Clustering(谱聚类)是一种基于图论的聚类方法,它能够识别任意形状的样本空间且收敛于全局最有解,其基本思想是利用样本数据的相似矩阵进行特征分解后得到的特征向量进行聚类,可见,它与样本feature无关而只与样本个数有关。

谱聚类(Spectral Clustering, SC)是一种基于图论的聚类方法——将带权无向图划分为两个或两个以上的最优子图,使子图内部尽量相似,而子图间距离尽量距离较远,以达到常见的聚类的目的。其中的最优是指最优目标函数不同,可以是割边最小分割——如图1的Smallest cut(如后文的Min cut), 也可以是分割规模差不多且割边最小的分割——如图1的Best cut(如后文的Normalized cut)。这样,谱聚类能够识别任意形状的样本空间且收敛于全局最优解,其基本思想是利用样本数据的相似矩阵(拉普拉斯矩阵)进行特征分解后得到的特征向量进行聚类。



Fast and efficient spectral clustering

SpectralClustering performs one of three spectral clustering algorithms (Unnormalized, Shi & Malik, Jordan & Weiss) on a given adjacency matrix. SimGraph creates such a matrix out of a given set of data and a given distance function.



UPDATE 09/13/2012


This major update to the final version includes

[+] Full GUI

[+] Several Plot Options: 2D/3D, Star Coordinates, Matrix Plot

[+] Save Plots

[+] Save and Load all kind of data (pure data, similarity graph, clustered data)

[+] Differentiates between already labeled and unlabeled data (see README).



The code has been optimized (within Matlab) to be both fast and memory efficient. Please look into the files and the Readme.txt for further information.



- Ulrike von Luxburg, "A Tutorial on Spectral Clustering", Statistics and Computing 17 (4), 2007


If there are any questions or suggestions, I will gladly help out. Just contact me at admin (at) airblader (dot) de

Bachelorarbeit.pdf  Main.m  CreateDataset.m  CreateDataset2.m  SpectralClustering.m  guiMain.m  setupGUI.m  convertClusterVector.m  normalizeData.m  openPlotFigure.m  
标签: 谱聚类 


用户反馈   关于我们
Copyright (©) ZHIHUISHI.COM 2013 All Rights Reserved.