Current browse context:
cs.LG
Change to browse by:
References & Citations
Computer Science > Machine Learning
Title: A unified framework for hard and soft clustering with regularized optimal transport
(Submitted on 12 Nov 2017 (v1), last revised 7 Mar 2024 (this version, v2))
Abstract: In this paper, we formulate the problem of inferring a Finite Mixture Model from discrete data as an optimal transport problem with entropic regularization of parameter $\lambda\geq 0$. Our method unifies hard and soft clustering, the Expectation-Maximization (EM) algorithm being exactly recovered for $\lambda=1$. The family of clustering algorithm we propose rely on the resolution of nonconvex problems using alternating minimization. We study the convergence property of our generalized $\lambda-$EM algorithms and show that each step in the minimization process has a closed form solution when inferring finite mixture models of exponential families. Experiments highlight the benefits of taking a parameter $\lambda>1$ to improve the inference performance and $\lambda\to 0$ for classification.
Submission history
From: Nicolas Papadakis [view email][v1] Sun, 12 Nov 2017 21:52:54 GMT (18kb)
[v2] Thu, 7 Mar 2024 20:43:31 GMT (1765kb,D)
Link back to: arXiv, form interface, contact.