Abstract: Spectral clustering is one of the most popular and important clustering methods in pattern recognition, machine learning, and data mining. However, its high computational complexity limits it in applications involving truly large-scale datasets. For a clustering problem with n samples, it needs to compute the eigenvectors of the graph Laplacian with O(n³) time complexity. To address this problem, we propose a novel method called anchor-based spectral clustering (ASC) by employing anchor points of data. Specifically, m (m≪n) anchor points are selected from the dataset, which can basically maintain the intrinsic (manifold) structure of the original data. Then a mapping matrix between the original data and the anchors is constructed. More importantly, it is proved that this data-anchor mapping matrix essentially preserves the clustering structure of the data. Based on this mapping matrix, it is easy to approximate the spectral embedding of the original data. The proposed method scales linearly relative to the size of the data but with low degradation of the clustering performance. The proposed method, ASC, is compared to the classical spectral clustering and two state-of-the-art accelerating methods, i.e., power iteration clustering and landmark-based spectral clustering, on 10 real-world applications under three evaluation metrics. Experimental results show that ASC is consistently faster than the classical spectral clustering with comparable clustering performance, and at least comparable with or better than the state-of-the-art methods on both effectiveness and efficiency.

一种基于锚点的谱聚类方法

[1]Arthur D, Vassilvitskii S, 2007. K-means++: the advantages of careful seeding. 18^th Annual ACM-SIAM Symp on Discrete Algorithms, p.1027-1035.

[2]Boutsidis C, Gittens A, Kambadur P, 2015. Spectral clustering via the power method–-provably. 32^nd Int Conf on Machine Learning, p.40-48.

[3]Chang XJ, Nie FP, Ma ZG, et al., 2015. A convex formulation for spectral shrunk clustering. 29^th AAAI Conf on Artificial Intelligence, p.2532-2538.

[4]Chen WY, Song YQ, Bai HJ, et al., 2011. Parallel spectral clustering in distributed systems. IEEE Trans Patt Anal Mach Intell, 33(3):568-586.

[5]Chen XL, Cai D, 2011. Large scale spectral clustering with landmark-based representation. 25^th AAAI Conf on Artificial Intelligence, p.313-318.

[6]Davies DL, Bouldin DW, 1979. A cluster separation measure. IEEE Trans Patt Anal Mach Intell, 1(2):224-227.

[7]Delalleau O, Bengio Y, Le Roux N, 2005. Efficient non-parametric function induction in semi-supervised learning. 10^th Int Workshop on Artificial Intelligence and Statistics, p.96-103.

[8]Demvšar J, 2006. Statistical comparisons of classifiers over multiple data sets. J Mach Learn Res, 7:1-30.

[9]Fowlkes C, Belongie S, Chung F, et al., 2004. Spectral grouping using the Nyström method. IEEE Trans Patt Anal Mach Intell, 26(2):214-225.

[10]Jia HJ, Ding SF, Xu XZ, et al., 2014. The latest research progress on spectral clustering. Neur Comput Appl, 24(7-8):1477-1486.

[11]Li HZ, Hu XG, Lin YJ, et al., 2016. A social tag clustering method based on common co-occurrence group similarity. Front Inform Technol Electron Eng, 17(2):122-134.

[12]Li JY, Xia YJ, Shan ZY, et al., 2015. Scalable constrained spectral clustering. IEEE Trans Knowl Data Eng, 27(2):589-593.

[13]Lin F, Cohen WW, 2010. Power iteration clustering. 27^th Int Conf on Machine Learning, p.655-662.

[14]Liu JL, Wang C, Danilevsky M, et al., 2013. Large-scale spectral clustering on graphs. 23^rd Int Joint Conf on Artificial Intelligence, p.1486-1492.

[15]Liu W, He JF, Chang SF, 2010. Large graph construction for scalable semi-supervised learning. 27^th Int Conf on Machine Learning, p.679-686.

[16]Luo MN, Nie FP, Chang XJ, et al., 2017. Adaptive unsupervised feature selection with structure regularization. IEEE Trans Neur Netw Learn Syst, 29(4):944-956.

[17]Mall R, Langone R, Suykens JAK, 2013a. FURS: fast and unique representative subset selection retaining large-scale community structure. Soc Netw Anal Min, 3(4):1075-1095.

[18]Mall R, Langone R, Suykens JAK, 2013b. Kernel spectral clustering for big data networks. Entropy, 15(5):1567-1586.

[19]Mall R, Jumutc V, Langone R, et al., 2014. Representative subsets for big data learning using K-NN graphs. IEEE Int Conf on Big Data, p.37-42.

[20]Ng AY, Jordan MI, Weiss Y, et al., 2002. On spectral clustering: analysis and an algorithm. Advances in Neural Information Processing Systems, p.849-856.

[21]Shi JB, Malik J, 2000. Normalized cuts and image segmentation. IEEE Trans Patt Anal Mach Intell, 22(8):888-905.

[22]Song YQ, Chen WY, Bai HJ, et al., 2008. Parallel spectral clustering. In: Daelemans W, Goethals B, Morik K (Eds.), Machine Learning and Knowledge Discovery in Databases. Springer Berlin Heidelberg, p.374-389.

[23]Tian F, Gao B, Cui Q, et al., 2014. Learning deep representations for graph clustering. 28^th AAAI Conf on Artificial Intelligence, p.1293-1299.

[24]von Luxburg U, 2007. A tutorial on spectral clustering. Stat Comput, 17(4):395-416.

[25]Wang L, Leckie C, Ramamohanarao K, et al., 2009. Approximate spectral clustering. In: Theeramunkong T, Kijsirikul B, Cercone N, et al. (Eds.), Advances in Knowledge Discovery and Data Mining. Springer Berlin Heidelberg, p.134-146.

[26]Xia RK, Pan Y, Du L, et al., 2014. Robust multi-view spectral clustering via low-rank and sparse decomposition. 28^th AAAI Conf on Artificial Intelligence, p.2149-2155.

[27]Xiang T, Gong SG, 2008. Spectral clustering with eigenvector selection. Patt Recog, 41(3):1012-1029.

[28]Xiao P, Li ZY, Guo S, et al., 2016. A $K$ self-adaptive SDN controller placement for wide area networks. Front Inform Technol Electron Eng, 17(7):620-633.

[29]Yan DH, Huang L, Jordan MI, 2009. Fast approximate spectral clustering. 15^th Int Conf on Knowledge Discovery and Data Mining, p.907-916.

[30]Yang Y, Xu D, Nie FP, et al., 2010. Image clustering using local discriminant models and global integration. IEEE Trans Image Process, 19(10):2761-2773.

[31]Yang Y, Shen HT, Nie FP, et al., 2011. Nonnegative spectral clustering with discriminative regularization. 25^th AAAI Conf on Artificial Intelligence, p.555-560.

[32]Yang Y, Nie F, Xu D, et al., 2012. A multimedia retrieval framework based on semi-supervised ranking and relevance feedback. IEEE Trans Patt Anal Mach Intell, 34(4):723-742.

[33]Zhang XC, Zong LL, You QZ, et al., 2016. Sampling for Nystrom extension-based spectral clustering: incremental perspective and novel analysis. ACM Trans Knowl Discov Data, 11(1):1-25.

[35]Zhu XJ, Lafferty J, 2005. Harmonic mixtures: combining mixture models and graph-based methods for inductive and scalable semi-supervised learning. 22^nd Int Conf on Machine Learning, p.1052-1059.