Support Vector Clustering combined with Spectral Graph Partitioning
Jin-Hyeong Park, Xiang Ji, Hongyuan Zha, Rangachar Kasturi · 2004
In this paper, we propose a new support vector cluster-ing (SVC) strategy by combining (SVC) with spectral graph partitioning (SGP). SVC has two main steps: support vector computation and cluster labeling using adjacency matrix. Spectral graph partitioning (SGP) method is applied to the adjacency matrix to determine the cluster labels. It is feasi-ble to combine multiple adjacency matrices computed using different parameters. A novel multi-resolution combination method is proposed for cluster labeling using the SGP for the purpose of boosting the clustering performance. 1.