Abstract | ||
---|---|---|
We focus on the problem of correlation clustering, which is to partition data points into clusters so that the repulsion within one cluster and the attraction between clusters could be as small as possible without predefining the number of clusters k. Finding the optimal solution to the problem is proven to be NP-hard, and various algorithms have been proposed to solve the problem approximately. Unfortunately, most of them are incapable of handling large-scale data. In this paper, we relax the problem by decoupling the affinity matrix and cluster indicator matrix, and propose a pseudo-EM optimization method to improve the scalability. Experimental results on synthetic data and real world problems including image segmentation and community detection show that our technique achieves state of the art performance in terms of both accuracy and scalability. |
Year | DOI | Venue |
---|---|---|
2013 | 10.1007/978-3-642-53917-6_2 | ADMA (2) |
Keywords | Field | DocType |
correlation clustering,large scale,pseudo-em algorithm,unsupervised learning | Data point,k-medians clustering,Data mining,Fuzzy clustering,Correlation clustering,Computer science,Image segmentation,Synthetic data,Artificial intelligence,Cluster analysis,Machine learning,Scalability | Conference |
Volume | Issue | Citations |
8347 LNAI | PART 2 | 1 |
PageRank | References | Authors |
0.35 | 14 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Yubo Wang | 1 | 24 | 5.49 |
Linli Xu | 2 | 790 | 42.51 |
Yucheng Chen | 3 | 36 | 5.91 |
H. Wang | 4 | 665 | 49.59 |