Comparing high dimensional partitions with the Coclustering Adjusted Rand Index

Valérie Robert 1, 2 Yann Vasseur 3, 2 Vincent Brault 4
3 SELECT - Model selection in statistical learning
Inria Saclay - Ile de France, LMO - Laboratoire de Mathématiques d'Orsay, CNRS - Centre National de la Recherche Scientifique : UMR
4 SVH - Statistique pour le Vivant et l’Homme
LJK - Laboratoire Jean Kuntzmann
Abstract : The popular Adjusted Rand Index (ARI) is extended to the task of simultaneous clustering of the rows and columns of a given matrix. This new index called Coclustering Adjusted Rand Index (CARI) remains convenient and competitive facing other indices. Indeed, partitions with high numbers of clusters can be considered and it does not require any convention when the numbers of clusters in partitions are different. Experiments on simulated partitions are presented and the performance of this index to measure the agreement between two pairs of partitions is assessed. Comparison with other indices is discussed.
Document type :
Preprints, Working Papers, ...
Complete list of metadatas

Cited literature [18 references]  Display  Hide  Download

https://hal.inria.fr/hal-01524832
Contributor : Valérie Robert <>
Submitted on : Monday, January 14, 2019 - 1:07:59 PM
Last modification on : Monday, September 2, 2019 - 9:42:03 AM

File

article_CARI.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01524832, version 4

Citation

Valérie Robert, Yann Vasseur, Vincent Brault. Comparing high dimensional partitions with the Coclustering Adjusted Rand Index. 2019. ⟨hal-01524832v4⟩

Share

Metrics

Record views

207

Files downloads

758