A Riemannian Optimization Approach to Clustering Problems

arXiv2208.03858MaRDI QIDQ6407232FDOQ6407232

Authors: Wen Huang, Meng Wei, K. A. Gallivan, Paul van Dooren

Publication date: 7 August 2022

Abstract: This paper considers the optimization problem in the form of

m i n_{X i n m a t h c a l F_{v}} f (x) + l a m b d a | X |_{1},

where

f

is smooth,

m a t h c a l F_{v} = X i n m a t h b b R^{n i m e s q} : X^{T} X = I_{q}, v i n m a t h r m s p a n (X)

, and

v

is a given positive vector. The clustering models including but not limited to the models used by

k

-means, community detection, and normalized cut can be reformulated as such optimization problems. It is proven that the domain

m a t h c a l F_{v}

forms a compact embedded submanifold of

m a t h b b R^{n i m e s q}

and optimization-related tools are derived. An inexact accelerated Riemannian proximal gradient method is proposed and its global convergence is established. Numerical experiments on community detection in networks and normalized cut for image segmentation are used to demonstrate the performance of the proposed method.

This page was built for publication: A Riemannian Optimization Approach to Clustering Problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6407232)