Concept Decompositions for Large Sparse Text Data Using Clustering

By Inderjit S. Dhillon, Dharmendra S. Modha. Machine-mediated learning, 2004

1,567 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computational Theory and Mathematics, Information Systems

View Paper: https://www.semanticscholar.org/paper/155d81d9f764de34c7234f0e3dc37a5bd297edee

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computational Theory and Mathematics

In Information Systems

Citations in the corpus

Scalable clustering of news search results (2011, neutral): “Other research related to clustering the search results focus on utilizing various matrix factorization techniques [15, 14] such as singular value decomposition, non-negative matrix factorization [12], local non-negative matrix factorization [13] and concept decomposition [4].”
A Divisive Information-Theoretic Feature Clustering Algorithm for Text Classification (2003, methodology): “Our initialization strategy is important, see step 1 in Figure 1 (a similar strategy was used by Dhillon and Modha, 2001, Section 5.1, to obtain word clusters), since it guarantees that the support set of everyp(C|wt) is contained in the support set of at least one cluster distribution p(C|W j), i.e”
Learning Feature Representations with K-Means (2012, methodology): “In this chapter, we will use a modiﬁed version of K-means (sometimes called “gain shape” vector quantization [41], or “spherical K-means” [14]) that ﬁnds D according to: minimize D ,s i and |”
A data driven method for feature transformation (2012, methodology): “Dot product as the similarity measure is frequently utilized in the document retrieval research under the term ”cosine similarity” and is tied naturally to the popular variant of the k-means algorithm known as spherical k-means [9], which groups points based on their cosine similarity.”
Localized matrix factorization for recommendation based on matrix block diagonal forms (2013, background): “Most efforts to improve system scalability focus on matrix clustering techniques [38, 31, 10, 9, 37] or designing incremental and distributed versions of existing MF algorithms [30, 20, 11].”
Generalized Low Rank Approximations of Matrices (2004, background): “…dimen-This The problem of low rank approximations of matrices has recently received broad attention in areas such as computer vision, information retrieval, and machine learning (Berry et al., 1995; Castelli et al., 2003; Deer-wester et al., 1990; Dhillon & Modha, 2001; Srebro & Jaakkola, 2003).”
Empirical and Theoretical Comparisons of Selected Criterion Functions for Document Clustering (2004, methodology): “The I2 criterion function (Eq. (2)) is used by the popular vector-space variant of the K -means algorithm (Cutting et al., 1992; Larsen & Aone, 1999; Dhillon & Modha, 2001; Steinbach, Karypis, & Kumar, 2000).”
Constrained Semi-supervised Learning in the Presence of Unanticipated Classes (2016, methodology): “The spherical K-Means algorithm [37], i.”
Active Semi-Supervision for Pairwise Constrained Clustering (2004, methodology): “Another dataset we used in our experiments is a subset of Classic3 [11] containing 400 documents – 100 Cranfield documents from aeronautical system papers, 100 Medline documents from medical journals , and 200 Cisi documents from information retrieval papers .”
Hidden Markov Model on a unit hypersphere space for gesture trajectory recognition (2014, methodology): “In the case of unit-hypersphere space data, Dhillon and Modha (2001) proposed a spherical k-means algorithm using cosine similarity instead of tandem Euclidian distance measure.”
Classifier ensembles: Select real-world applications (2008, background): “In general however, there can be many different clusterings for the same data, depending on the clustering algorithm used [22,58,62].”
Clustering on the Unit Hypersphere using von Mises-Fisher Distributions (2005, methodology): “Further, the spherical kmeans (spkmeans) algorithm (Dhillon and Modha, 2001), that performs kmeans using cosine similarity instead of Euclidean distortion, has been found to work well for text clustering.”
A probabilistic framework for semi-supervised clustering (2004, methodology): “Since constraints do not take part in cluster representative reestimation, this step remains the same as in K-Means for Bregman divergences, and the same as in SPKMEANS for weighted cosine similarity [18].”