An Exact Relaxation of Clustering

Morten Mørup, Lars Kai Hansen

AbstractContinuous relaxation of hard assignment clustering problems can lead to better solutions than greedy iterative refinement algorithms. However, the validity of existing relaxations is contingent on problem specific fuzzy parameters that quantify the level of similarity between the original combinatorial problem and the relaxed continuous domain problem. Equivalence of solutions obtained from the relaxation and the hard assignment is guaranteed only in the limit of vanishing `fuzzyness'.
This paper derives a new exact relaxation without such a fuzzy parameter which is applicable for a wide range of clustering problems such as the k-means objective and pairwise clustering as well as
graph partition problems, e.g., for community detection in complex networks. In particular we show that a relaxation to the simplex can be given for which the extreme solutions are stable hard assignment solutions and vice versa. Based on the new relaxation we derive the
sr-clustering algorithm that has the same complexity as traditional greedy iterative refinement algorithms but leading to significantly better partitions of the data. A Matlab implementation of the sr-clustering algorithm is available for download.
KeywordsClustering, Complex Networks, Community Detection, Simplicial Relaxation, sr-clustering
TypeTechnical report
Electronic version(s)[pdf]
BibTeX data [bibtex]
IMM Group(s)Intelligent Signal Processing

Back  ::  IMM Publications