Abstract | ||
---|---|---|
As an important unsupervised learning approach, clustering is widely used in pattern recognition, information retrieval and image analysis, etc. In various clustering approaches, graph based clustering has received much interest and obtain impressive success in application recently. However, existing graph based clustering algorithms usually require as input some parameters in one form or another. In this paper we study the dominant sets clustering algorithm and present a new clustering algorithm without any parameter input. We firstly use histogram equalization to transform the similarity matrices of data. This transformation is shown to make the clustering results invariant to similarity parameters effectively. Then we merge clusters based on the ratio between intra-cluster and inter-cluster similarity. Our algorithm is shown to be effective in experiments on seven datasets. |
Year | DOI | Venue |
---|---|---|
2015 | 10.1007/978-3-319-24261-3_8 | Lecture Notes in Computer Science |
Field | DocType | Volume |
Fuzzy clustering,Canopy clustering algorithm,Data mining,CURE data clustering algorithm,Data stream clustering,Pattern recognition,Correlation clustering,Artificial intelligence,Constrained clustering,Cluster analysis,Mathematics,Single-linkage clustering | Conference | 9370 |
ISSN | Citations | PageRank |
0302-9743 | 1 | 0.35 |
References | Authors | |
16 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Jian Hou | 1 | 18 | 3.08 |
Chunshi Sha | 2 | 1 | 0.35 |
Hongxia Cui | 3 | 24 | 4.90 |
Lei Chi | 4 | 1 | 0.69 |