We present a new clustering algorithm by proposing a convex relaxation of hierarchical clustering, which results in a family of objective functions with a natural geometric interpretation. We give efficient algorithms for calculating the continuous regularization path of solutions, and discuss relative advantages of the parameters. Our method experimentally gives state-of-the-art results similar to spectral clustering for non-convex clusters, and has the added benefit of learning a tree structure from the data.
Publié le : 2011-06-27
Classification:
Clustering,
Sparsity,
Fusion,
Hierarchical,
Convex,
Optimization,
Unsupervised,
Learning,
[STAT.ML]Statistics [stat]/Machine Learning [stat.ML],
[MATH.MATH-OC]Mathematics [math]/Optimization and Control [math.OC],
[INFO.INFO-LG]Computer Science [cs]/Machine Learning [cs.LG]
@article{hal-00591630,
author = {Hocking, Toby Dylan and Joulin, Armand and Bach, Francis and Vert, Jean-Philippe},
title = {Clusterpath An Algorithm for Clustering using Convex Fusion Penalties},
journal = {HAL},
volume = {2011},
number = {0},
year = {2011},
language = {en},
url = {http://dml.mathdoc.fr/item/hal-00591630}
}
Hocking, Toby Dylan; Joulin, Armand; Bach, Francis; Vert, Jean-Philippe. Clusterpath An Algorithm for Clustering using Convex Fusion Penalties. HAL, Tome 2011 (2011) no. 0, . http://gdmltest.u-ga.fr/item/hal-00591630/