Kernel Clustering: Density Biases and solutions

被引:35
|
作者
Marin, Dmitrii [1 ]
Tang, Meng [1 ]
Ben Ayed, Ismail [2 ]
Boykov, Yuri [1 ]
机构
[1] Univ Western Ontario, Dept Comp Sci, London, ON N6A 3K7, Canada
[2] Univ Quebec, Ecole Technol Super, Mont Royal, PQ H3R 1K, Canada
基金
加拿大自然科学与工程研究理事会;
关键词
Kernel methods; kernel clustering; kernel k-means; average association; average cut; normalized cut; dominant set;
D O I
10.1109/TPAMI.2017.2780166
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Kernel methods are popular in clustering due to their generality and discriminating power. However, we show that many kernel clustering criteria have density biases theoretically explaining some practically significant artifacts empirically observed in the past. For example, we provide conditions and formally prove the density mode isolation bias in kernel K-means for a common class of kernels. We call it Breiman's bias due to its similarity to the histogram mode isolation previously discovered by Breiman in decision tree learning with Gini impurity. We also extend our analysis to other popular kernel clustering methods, e.g., average/normalized cut or dominant sets, where density biases can take different forms. For example, splitting isolated points by cut-based criteria is essentially the sparsest subset bias, which is the opposite of the density mode bias. Our findings suggest that a principled solution for density biases in kernel clustering should directly address data inhomogeneity. We show that density equalization can be implicitly achieved using either locally adaptive weights or locally adaptive kernels. Moreover, density equalization makes many popular kernel clustering objectives equivalent. Our synthetic and real data experiments illustrate density biases and proposed solutions. We anticipate that theoretical understanding of kernel clustering limitations and their principled solutions will be important for a broad spectrum of data analysis applications across the disciplines.
引用
收藏
页码:136 / 147
页数:12
相关论文
共 50 条
  • [21] A new measure for assessment of clustering based on kernel density estimation
    Modak, Soumita
    COMMUNICATIONS IN STATISTICS-THEORY AND METHODS, 2023, 52 (17) : 5942 - 5951
  • [22] An Improved Fast Search Clustering Algorithm Based on Kernel Density
    Zhang, Ruisheng
    Ma, Huiyi
    Liu, Qidong
    Zhao, Zhili
    2015 IEEE INTERNATIONAL CONFERENCE ON SMART CITY/SOCIALCOM/SUSTAINCOM (SMARTCITY), 2015, : 689 - 693
  • [23] Density peaks clustering algorithm based on kernel density estimation and minimum spanning tree
    Fan T.
    Li X.
    Hou J.
    Liu B.
    Kang P.
    International Journal of Innovative Computing and Applications, 2022, 13 (5-6) : 336 - 350
  • [24] CHALLENGES AND POSSIBLE SOLUTIONS TO DENSITY BASED CLUSTERING
    Yasar, Fatma Gunseli
    Ulutagay, Gozde
    2016 IEEE 8TH INTERNATIONAL CONFERENCE ON INTELLIGENT SYSTEMS (IS), 2016, : 492 - 498
  • [25] On Potts model clustering, kernel K-means, and density estimation
    Murua, Alejandro
    Stanberry, Larissa
    Stuetzle, Werner
    JOURNAL OF COMPUTATIONAL AND GRAPHICAL STATISTICS, 2008, 17 (03) : 629 - 658
  • [26] MulticlusterKDE: a new algorithm for clustering based on multivariate kernel density estimation
    Scaldelai, D.
    Matioli, L. C.
    Santos, S. R.
    Kleina, M.
    JOURNAL OF APPLIED STATISTICS, 2022, 49 (01) : 98 - 121
  • [27] Object detection by clustering-based nonparametric kernel density estimation
    Hu, D.
    Hu, J.
    INFORMATION SCIENCE AND MANAGEMENT ENGINEERING, VOLS 1-3, 2014, 46 : 1867 - 1872
  • [28] Fuzzy Models Synthesis with Kernel-Density-Based Clustering Algorithm
    Lukasik, Szymon
    Kowalski, Piotr A.
    Charytanowicz, Malgorzata
    Kulczycki, Piotr
    FIFTH INTERNATIONAL CONFERENCE ON FUZZY SYSTEMS AND KNOWLEDGE DISCOVERY, VOL 3, PROCEEDINGS, 2008, : 449 - 453
  • [29] Density-sensitive fuzzy kernel maximum entropy clustering algorithm
    Li Y.-T.
    Guo J.
    Qi L.
    Liu X.
    Ruan P.-Y.
    Tao X.-M.
    Kongzhi Lilun Yu Yingyong/Control Theory and Applications, 2022, 39 (01): : 67 - 82
  • [30] Kernel density estimation, affinity-based clustering, and typical cuts
    Erdogmus, Deniz
    Carreira-Perpinan, Miguel A.
    Ozertem, Umut
    2006 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING, VOLS 1-13, 2006, : 5427 - 5430