DMCA
Computational Aspects of Lucidity-Driven Graph Clustering (2010)
Cached
Download Links
| Citations: | 2 - 1 self |
BibTeX
@MISC{Görke10computationalaspects,
author = {Robert Görke and Marco Gaertler and Florian Hübner and Dorothea Wagner},
title = {Computational Aspects of Lucidity-Driven Graph Clustering},
year = {2010}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
We formally state and investigate the lucidity paradigm for graph clusterings. The rationale that substantiates this concept is the trade-off between the achieved quality and the expected quality of a graph clustering. A recently defined quality measure for graph clusterings, modularity, is one specific realization of this paradigm, in fact the pioneering one. On a theoretical side, we state a sound probability space for lucidity and thus for modularity and prove that in this paradigm of lucidity, using a subtractive trade-off and either the index coverage (yields modularity) or performance leads to equivalent indices. Furthermore, we show that the NP-hardness of optimizing these indices yields the NP-hardness of the problem MinMixedMultiPartition. Moreover, we describe an efficient maximization algorithm for a divisive trade-off between quality and expectation. We experimentally evaluate four realizations of this paradigm systematically and confirm their feasibility in a first methodic analysis of the behavior of these realizations on both artificial and on real-world data, arriving at good results of community assessment and detection.
Keyphrases
computational aspect lucidity-driven graph clustering graph clustering specific realization sound probability space first methodic analysis index coverage subtractive trade-off theoretical side efficient maximization algorithm good result problem minmixedmultipartition divisive trade-off real-world data performance lead lucidity paradigm community assessment equivalent index yield modularity expected quality quality measure
