Clustering very large data sets with principal direction divisive partitioning

D. Littau, D. Boley

Research output: Chapter in Book/Report/Conference proceedingChapter

11 Scopus citations


We present a method to cluster data sets too large to fit in memory, based on a Low-Memory Factored Representation (LMFR). The LMFR represents the original data in a factored form with much less memory, while preserving the individuality of each of the original samples. The scalable clustering algorithm Principal Direction Divisive Partitioning (PDDP) can use the factored form in a natural way to obtain a clustering of the original dataset. The resulting algorithm is the PieceMeal PDDP (PMPDDP) method. The scalability of PMPDDP is demonstrated with a complexity analysis and experimental results. A discussion on the practical use of this method by a casual user is provided.

Original languageEnglish (US)
Title of host publicationGrouping Multidimensional Data
Subtitle of host publicationRecent Advances in Clustering
PublisherSpringer Berlin Heidelberg
Number of pages28
ISBN (Print)354028348X, 9783540283485
StatePublished - 2006


Dive into the research topics of 'Clustering very large data sets with principal direction divisive partitioning'. Together they form a unique fingerprint.

Cite this