文件列表:

Demo.m (2276, 2012-01-03)
SimGraph.m (7888, 2012-01-03)
SpectralClustering.m (2552, 2012-01-03)
distEuclidean.m (391, 2012-01-02)
generatedata (0, 2012-01-04)
generatedata\GenData_Ellipse.m (1466, 2012-01-02)
generatedata\GenData_Gaussian.m (793, 2012-01-03)
license.txt (1331, 2012-01-03)

% =========================== % Bachelor Thesis % Author : Ingo Bürk % Year : 2011/2012 % Contact: admin@airblader.de % =========================== % =========================================================== % GENERAL INFORMATION % =========================================================== The Matlab files provided can create sample data, different kinds of similarity graphs and perform fast and efficient spectral clustering algorithms. If you need help, please read this file first and try typing 'help [Filename]' to get information. If there are still questions left, I'll be happy to help you upon contacting me via email. % =========================================================== % TECHNICAL INFORMATION % =========================================================== If you load and use your own data (adjacency matrix), please keep in mind that using a sparse matrix will reduce memory drastically. All methods in these files work with sparse matrices and therefore assume a sparse structure. If your data is not sparse per se, consider using a similarity graph that will give it a sparse structure. We recommend using either an epsilon or mutual k-Nearest neighbors similarity graph, combined with the spectral clustering algorithm according to Shi and Malik (2000). % =========================================================== % FILES % =========================================================== - SpectralClustering.m This method will perform one of three spectral clustering algorithms on a given adjacency matrix. Please type 'help SpectralClustering' for further information. - SimGraph.m With this method, you can create one of several kinds of similarity graphs out of a given set of data or a given distance matrix. This file has been optimized to be both fast and memory efficient. - distEuclidean.m Calculated Euclidean distances between two matrices containing data points. This is the default method that will be used when creating a similarity graph out of a given set of data. Please look into this file in order to get information on how to write your own distance function. - generatedata/GenData_Ellipse.m This function returns a sample set of two-dimensional data in the shape of an ellipse with given parameters. It is for testing purposes only. - generatedata/GenData_Gaussian.m This function returns a sample set of data in the shape of a Gaussian distributed data cloud with given parameters. It is for testing purposes only. % =========================================================== % REFERENCES % =========================================================== This thesis and the resulting work is based on - Ulrike von Luxburg, "A Tutorial on Spectral Clustering", Statistics and Computing 17 (4), 2007

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：