Bilkent University
Department of Computer Engineering
S E M I N A R
Cluster Ensemble Framework for Hypergraph Partitioning
Onur Özbek
MSc Student
Computer Engineering Department
Bilkent UniversityHypergraph partitioning has been recognized as an important problem and has wide applicability in many areas. The problem is to divide a hypergraph into two or more roughly equal sized parts while minimizing the number of hyperedges connecting different parts, and several heuristic algorithms have been developed. Cluster ensemble methods aim to generate a more accurate clustering of data by using the results from different clustering algorithms. We propose a new method which combines these two approaches in to create an optimal partitioning of an hypergraph. First, an existing partitioning algorithm is used on slightly-altered versions of the initial hypergraph. Then, a bipartite graph is constructed with edges establishing membership between vertices of the hypergraph and partitions acquired at the previous step. Finally, the bipartite graph is partitioned to obtain the final result.
Keywords: cluster ensemble, hypergraph partitioning
DATE: 07 May, 2012, Monday @ 16:20
PLACE: EA409