Fast affinity propagation clustering: A multilevel approach

• Published in: Pattern recognition Vol. 45; no. 1; pp. 474 - 486
• Main Authors: Shang, Fanhua, Jiao, L.C., Shi, Jiarong, Wang, Fei, Gong, Maoguo
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 2012; Elsevier
• Subjects: Affinity propagation; Applied sciences; Clustering; Exact sciences and technology; Graph partitioning; Information, signal and communications theory; Manifold structure; Sampling, quantization; Signal and communications theory; Signal representation. Spectral analysis; Signal, noise; Spectral clustering; Telecommunications and information theory; Graph partitioning; Manifold structure; Spectral clustering; Clustering; Affinity propagation; Performance evaluation; Automatic classification; State of the art; Spectral method; Graph theory; Signal classification; Implementation; Graph method; Fast algorithm; Sampling; Domain decomposition; Multilevel system
• ISSN: 0031-3203; 1873-5142
• DOI: 10.1016/j.patcog.2011.04.032

In this paper, we propose a novel Fast Affinity Propagation clustering approach ( FAP). FAP simultaneously considers both local and global structure information contained in datasets, and is a high-quality multilevel graph partitioning method that can implement both vector-based and graph-based clustering. First, a new Fast Sampling algorithm ( FS) is proposed to coarsen the input sparse graph and choose a small number of final representative exemplars. Then a density-weighted spectral clustering method is presented to partition those exemplars on the global underlying structure of data manifold. Finally, the cluster assignments of all data points can be achieved through their corresponding representative exemplars. Experimental results on two synthetic datasets and many real-world datasets show that our algorithm outperforms the state-of-the-art original affinity propagation and spectral clustering algorithms in terms of speed, memory usage, and quality on both vector-based and graph-based clustering. ► We present a global distance that is very robust against the noise and outliers. ► We propose a new fast sampling algorithm to identify representative exemplars. ► We propose a novel multilevel fast affinity propagation clustering approach.