References & Citations
Statistics > Machine Learning
Title: Co-clustering through Optimal Transport
(Submitted on 17 May 2017 (v1), last revised 19 May 2017 (this version, v3))
Abstract: In this paper, we present a novel method for co-clustering, an unsupervised learning approach that aims at discovering homogeneous groups of data instances and features by grouping them simultaneously. The proposed method uses the entropy regularized optimal transport between empirical measures defined on data instances and features in order to obtain an estimated joint probability density function represented by the optimal coupling matrix. This matrix is further factorized to obtain the induced row and columns partitions using multiscale representations approach. To justify our method theoretically, we show how the solution of the regularized optimal transport can be seen from the variational inference perspective thus motivating its use for co-clustering. The algorithm derived for the proposed method and its kernelized version based on the notion of Gromov-Wasserstein distance are fast, accurate and can determine automatically the number of both row and column clusters. These features are vividly demonstrated through extensive experimental evaluations.
Submission history
From: Charlotte Laclau [view email][v1] Wed, 17 May 2017 14:46:12 GMT (449kb,D)
[v2] Thu, 18 May 2017 05:49:49 GMT (449kb,D)
[v3] Fri, 19 May 2017 05:47:35 GMT (449kb,D)
Link back to: arXiv, form interface, contact.