Fast Discrete Distribution Clustering Using Wasserstein Barycenter With Sparse Support

被引:74
作者
Ye, Jianbo [1 ]
Wu, Panruo [2 ]
Wang, James Z. [1 ]
Li, Jia [3 ]
机构
[1] Penn State Univ, Coll Informat Sci & Technol, University Pk, PA 16802 USA
[2] Univ Calif Riverside, Dept Comp Sci & Engn, Riverside, CA 92521 USA
[3] Penn State Univ, Dept Stat, University Pk, PA 16802 USA
基金
美国国家科学基金会;
关键词
Discrete distribution; K-means; clustering; large-scale learning; parallel computing; ADMM; EARTH MOVERS DISTANCE;
D O I
10.1109/TSP.2017.2659647
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
In a variety of research areas, the weighted bag of vectors and the histogram are widely used descriptors for complex objects. Both can be expressed as discrete distributions. D2-clustering pursues the minimum total within-cluster variation for a set of discrete distributions subject to the Kantorovich-Wasserstein metric. D2-clustering has a severe scalability issue, the bottleneck being the computation of a centroid distribution, called Wasserstein barycenter, that minimizes its sum of squared distances to the cluster members. In this paper, we develop a modified Bregman ADMM approach for computing the approximate discrete Wasserstein barycenter of large clusters. In the case when the support points of the barycenters are unknown and have low cardinality, our method achieves high accuracy empirically at a much reduced computational cost. The strengths and weaknesses of our method and its alternatives are examined through experiments, and we recommend scenarios for their respective usage. Moreover, we develop both serial and parallelized versions of the algorithm. By experimenting with large-scale data, we demonstrate the computational efficiency of the new methods and investigate their convergence properties and numerical stability. The clustering results obtained on several datasets in different domains are highly competitive in comparison with some widely used methods in the corresponding areas.
引用
收藏
页码:2317 / 2332
页数:16
相关论文
共 50 条
  • [31] Clustering-based hyperspectral band selection using sparse nonnegative matrix factorization
    Ji-ming Li
    Yun-tao Qian
    Journal of Zhejiang University SCIENCE C, 2011, 12 : 542 - 549
  • [32] Clustering-based hyperspectral band selection using sparse nonnegative matrix factorization
    Li, Ji-ming
    Qian, Yun-tao
    JOURNAL OF ZHEJIANG UNIVERSITY-SCIENCE C-COMPUTERS & ELECTRONICS, 2011, 12 (07): : 542 - 549
  • [33] Fast agglomerative clustering using a k-nearest neighbor graph
    Franti, Pasi
    Virmajoki, Olli
    Hautamaki, Ville
    IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2006, 28 (11) : 1875 - 1881
  • [34] Fast Nearest Neighbor classification using class-based clustering
    Chen, Tung-Shou
    Chiu, Yung-Hsing
    Lin, Chih-Chiang
    PROCEEDINGS OF 2007 INTERNATIONAL CONFERENCE ON MACHINE LEARNING AND CYBERNETICS, VOLS 1-7, 2007, : 1894 - +
  • [35] Histological image segmentation using fast mean shift clustering method
    Wu, Geming
    Zhao, Xinyan
    Luo, Shuqian
    Shi, Hongli
    BIOMEDICAL ENGINEERING ONLINE, 2015, 14
  • [36] Discrete event estimation of nonlinear hybrid dynamical systems using fuzzy clustering
    Shah, Ankit K.
    Adhyaru, Dipak M.
    International Journal of Modelling and Simulation, 2015, 35 (3-4) : 137 - 142
  • [37] Segmentation of magnetic resonance images using discrete curve evolution and fuzzy clustering
    Supot, Sookpotharom
    Thanapong, Chaichana
    Chuchar, Pintavirooj
    Manas, Sangworasil
    2007 IEEE INTERNATIONAL CONFERENCE ON INTEGRATION TECHNOLOGY, PROCEEDINGS, 2007, : 697 - +
  • [38] Histological image segmentation using fast mean shift clustering method
    Geming Wu
    Xinyan Zhao
    Shuqian Luo
    Hongli Shi
    BioMedical Engineering OnLine, 14
  • [39] Robust clustering and outlier rejection using the Mahalanobis distance distribution
    Roizman, Violeta
    Jonckheere, Matthieu
    Pascal, Frederic
    28TH EUROPEAN SIGNAL PROCESSING CONFERENCE (EUSIPCO 2020), 2021, : 2448 - 2452
  • [40] Leak identification in a water distribution network using sparse flow measurements
    Mulholland, Michael
    Purdon, Andrew
    Latifi, M. Abderrazak
    Brouckaert, Christopher
    Buckley, Christopher
    COMPUTERS & CHEMICAL ENGINEERING, 2014, 66 : 252 - 258