A Dependence Maximization View of Clustering

Le Song, Alexander J. Smola, Arthur Gretton, Karsten M. Borgwardt

ICML 2007 pp. 815-822

doi:10.1145/1273496.1273599 /icml/2007/song2007icml-dependence/

Abstract

We propose a family of clustering algorithms based on the maximization of dependence between the input variables and their cluster labels, as expressed by the Hilbert-Schmidt Independence Criterion (HSIC). Under this framework, we unify the geometric, spectral, and statistical dependence views of clustering, and subsume many existing algorithms as special cases (e.g. k-means and spectral clustering). Distinctive to our framework is that kernels can also be applied on the labels, which can endow them with particular structures. We also obtain a perturbation bound on the change in k-means clustering.

PDF ICML Semantic Scholar

Cite

Text

Song et al. "A Dependence Maximization View of Clustering." International Conference on Machine Learning, 2007. doi:10.1145/1273496.1273599

Markdown

[Song et al. "A Dependence Maximization View of Clustering." International Conference on Machine Learning, 2007.](https://mlanthology.org/icml/2007/song2007icml-dependence/) doi:10.1145/1273496.1273599

BibTeX

@inproceedings{song2007icml-dependence,
  title     = {{A Dependence Maximization View of Clustering}},
  author    = {Song, Le and Smola, Alexander J. and Gretton, Arthur and Borgwardt, Karsten M.},
  booktitle = {International Conference on Machine Learning},
  year      = {2007},
  pages     = {815-822},
  doi       = {10.1145/1273496.1273599},
  url       = {https://mlanthology.org/icml/2007/song2007icml-dependence/}
}