Learning with symmetric positive definite matrices via generalized Bures-Wasserstein geometry

DOI10.1007/978-3-031-38271-0_40MaRDI QIDQ6138145zbMATH OpenOpenAlexFDO

Authors Bamdev Mishra, Pratik Jawanpuria, Junbin Gao

Publication date 16 January 2024

Published in Lecture Notes in Computer Science (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2110.10464

zbMATH Keywords

Bures-Wasserstein geometrysymmetric positive definite matrices

Mathematics Subject Classification ID

Differential geometric aspects of statistical manifolds and information geometry (53B12)

Abstract: Learning with symmetric positive definite (SPD) matrices has many applications in machine learning. Consequently, understanding the Riemannian geometry of SPD matrices has attracted much attention lately. A particular Riemannian geometry of interest is the recently proposed Bures-Wasserstein (BW) geometry which builds on the Wasserstein distance between the Gaussian densities. In this paper, we propose a novel generalization of the BW geometry, which we call the GBW geometry. The proposed generalization is parameterized by a symmetric positive definite matrix $mathbfM$ such that when $mathbfM=mathbfI$, we recover the BW geometry. We provide a rigorous treatment to study various differential geometric notions on the proposed novel generalized geometry which makes it amenable to various machine learning applications. We also present experiments that illustrate the efficacy of the proposed GBW geometry over the BW geometry.

Recommendations

• On the Bures-Wasserstein distance between positive definite matrices
• Curvature of the manifold of fixed-rank positive-semidefinite matrices endowed with the Bures-Wasserstein metric
• \(O(n)\)-invariant Riemannian metrics on SPD matrices
• A Riemannian submersion-based approach to the Wasserstein barycenter of positive definite matrices
• Differential geometry with extreme eigenvalues in the positive semidefinite cone

Cites work

• A Riemannian framework for tensor computing
• An alternative to EM for Gaussian mixture models: batch and stochastic Riemannian optimization
• An Introduction to Optimization on Smooth Manifolds
• Bures-Wasserstein geometry for positive-definite Hermitian matrices and their trace-one subset
• Computational optimal transport. With applications to data sciences
• Curvature of the manifold of fixed-rank positive-semidefinite matrices endowed with the Bures-Wasserstein metric
• scientific article; zbMATH DE number 5223994 (Why is no real title available?)
• Manopt, a Matlab toolbox for optimization on manifolds
• On the Bures-Wasserstein distance between positive definite matrices
• Positive Definite Matrices
• Positive definite matrices and the S-divergence
• Pymanopt: a Python toolbox for optimization on manifolds using automatic differentiation
• QUIC: quadratic approximation for sparse inverse covariance estimation
• Riemannian Geometry of Symmetric Positive Definite Matrices via Cholesky Decomposition
• Riemannian optimization with a preconditioning scheme on the generalized Stiefel manifold
• Riemannian preconditioning
• ROPTLIB: An object-oriented C++ library for optimization on Riemannian manifolds
• Sparse inverse covariance estimation with the graphical lasso
• Stochastic Gradient Descent on Riemannian Manifolds
• Wasserstein Riemannian geometry of Gaussian densities

Cited in

(1)

• Riemannian-geometric generalizations of quantum fidelities and Bures-Wasserstein distance

This page was built for publication: Learning with symmetric positive definite matrices via generalized Bures-Wasserstein geometry

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6138145)