Introduction to Clustering Large and High-Dimensional Data
Focusing on a few of the most important clustering algorithms, this book provides a detailed account of these major models in an information retrieval context.
TABLE OF CONTENTS Introduction to Clustering Large and High-Dimensional Data
2.5. Bibliographic Notes
2.5. Bibliographic Notes
For discussion of means see, for example, the classical monograph [67], and [13, 132]. The classical quadratic batch k-means algorithm is attributed to [57], the problem is mentioned already in the work of Hugo Steinhaus [125]. The incremental version of the algorithm is discussed in [47] and the combination of the two versions is described in [66, 78, 145]. The idea of changing cluster affiliation for a subset of the data set 
only as well as modification of the objective function that penalizes partitions with large number of clusters is also discussed in [78].
The compressed column storage format (CCS), which is also called the Harwell-Boeing sparse matrix format is given in [49]. For an objective function independent distance on the set of partitions of a given data set see [104].
Approximation of a vector set by a line leads to very fast clustering algorithm introduced in [20] (this algorithm is discussed in detail in Section 5.3.1). A centroid of a vector set ? R n that is an affine subspace of R n is discussed in [23], a representative of a set of categorical objects ( mode of a set ) is defined in [70]. Representation of a cluster by a solution of an optimization problem goes back to work of Diday (see, e.g., [42, 43]).
Computational shortcuts to accelerate the batch k-means algorithm with a distance like function d( x , y) that satisfies the triangle inequality are discussed...
Copyright Jacob Kogan 2007 under license agreement with Books24x7
