A unifying theorem for spectral embedding and clustering

@misc{citeulike:1395804, abstract = {Spectral methods use selected eigenvectors of a data affinity matrix to obtain a data representation that can be trivially clustered or embedded in a low-dimensional space. We present a theorem that explains, for broad classes of affinity matrices and eigenbases, why this works: For successively smaller eigenbases (i.e., using fewer and fewer of the affinity matrix's dominant eigenvalues and eigenvectors), the angles between \"similar\" vectors in the new representation shrink while the ...}, author = {Brand, M. and Huang, K. }, citeulike-article-id = {1395804}, keywords = {clustering, embedding, graph, spe, spectral, spectral-clustering}, posted-at = {2008-05-15 13:23:33}, priority = {0}, title = {A unifying theorem for spectral embedding and clustering}, url = {http://citeseer.ist.psu.edu/brand03unifying.html}, year = {2003} }

TY - GEN ID - citeulike:1395804 L3 - citeulike-article-id:1395804 TI - A unifying theorem for spectral embedding and clustering N2 - Spectral methods use selected eigenvectors of a data affinity matrix to obtain a data representation that can be trivially clustered or embedded in a low-dimensional space. We present a theorem that explains, for broad classes of affinity matrices and eigenbases, why this works: For successively smaller eigenbases (i.e., using fewer and fewer of the affinity matrix's dominant eigenvalues and eigenvectors), the angles between "similar" vectors in the new representation shrink while the ... KW - clustering KW - embedding KW - graph KW - spe KW - spectral KW - spectral-clustering AU - Brand, M AU - Huang, K PY - 2003/// UR - http://citeseer.ist.psu.edu/brand03unifying.html ER -