2010/11/17 by Andri Mirzal, Mirzal, Andri
Computer Science · #15A18 #65F15 #Data Mining Algorithms and Applications #FOS: Computer and information sciences #FOS: Mathematics #Machine Learning (cs.LG) #Neural Networks and Applications #Numerical Analysis (math.NA) #Rough Sets and Fuzzy Logic #Spectral Theory (math.SP)
paper · pdf · doi:10.48550/arxiv.1011.4104
openalex publication_date 2010/11/17 · openalex created_date 2022/10/03 · openalex updated_date 2026/07/28
This paper discusses clustering and latent semantic indexing (LSI) aspects of\nthe singular value decomposition (SVD). The purpose of this paper is twofold.\nThe first is to give an explanation on how and why the singular vectors can be\nused in clustering. And the second is to show that the two seemingly unrelated\nSVD aspects actually originate from the same source: related vertices tend to\nbe more clustered in the graph representation of lower rank approximate matrix\nusing the SVD than in the original semantic graph. Accordingly, the SVD can\nimprove retrieval performance of an information retrieval system since queries\nmade to the approximate matrix can retrieve more relevant documents and filter\nout more irrelevant documents than the same queries made to the original\nmatrix. By utilizing this fact, we will devise an LSI algorithm that mimicks\nSVD capability in clustering related vertices. Convergence analysis shows that\nthe algorithm is convergent and produces a unique solution for each input.\nExperimental results using some standard datasets in LSI research show that\nretrieval performances of the algorithm are comparable to the SVD's. In\naddition, the algorithm is more practical and easier to use because there is no\nneed to determine decomposition rank which is crucial in driving retrieval\nperformance of the SVD.\n