Clustering through the optimal transport barycenter problem
Yang, Hongkang ; Tabak, Esteban G.
arXiv, Tome 2019 (2019) no. 0, / Harvested from
The problem of clustering a data set is formulated in terms of the Wasserstein barycenter problem in optimal transport. The objective proposed is the maximization of the variability attributable to class, further characterized as the minimization of the variance of the Wasserstein barycenter. Existing theory, which constrains the transport maps to rigid translations, is generalized to include affine transformations, which are proved optimal for the purpose of clustering. The resulting non-parametric clustering algorithms include k-means as a special case and have more robust performance, demonstrated by comparisons with popular clustering algorithms on both artificial and real-world data sets.
Publié le : 2019-02-26
Classification:  Mathematics - Optimization and Control,  Mathematics - Statistics Theory,  62H30, 49M30 
@article{1902.10288,
     author = {Yang, Hongkang and Tabak, Esteban G.},
     title = {Clustering through the optimal transport barycenter problem},
     journal = {arXiv},
     volume = {2019},
     number = {0},
     year = {2019},
     language = {en},
     url = {http://dml.mathdoc.fr/item/1902.10288}
}

Yang, Hongkang; Tabak, Esteban G. Clustering through the optimal transport barycenter problem. arXiv, Tome 2019 (2019) no. 0, . http://gdmltest.u-ga.fr/item/1902.10288/