Conic geometric optimisation on the manifold of positive definite matrices

Abstract

We develop \emph{geometric optimisation} on the manifold of Hermitian positive definite (HPD) matrices. In particular, we consider optimising two types of cost functions: (i) geodesically convex (g-convex); and (ii) log-nonexpansive (LN). G-convex functions are nonconvex in the usual euclidean sense, but convex along the manifold and thus allow global optimisation. LN functions may fail to be even g-convex, but still remain globally optimisable due to their special structure. We develop theoretical tools to recognise and generate g-convex functions as well as cone theoretic fixed-point optimisation algorithms. We illustrate our techniques by applying them to maximum-likelihood parameter estimation for elliptically contoured distributions (a rich class that substantially generalises the multivariate normal distribution). We compare our fixed-point algorithms with sophisticated manifold optimisation methods and obtain notable speedups.

27 pages; updated version with simplified presentation; 7 figures

Cited by in corpus (43)

• Riemannian Adaptive Optimization Methods
• Parallel Transport on the Cone Manifold of SPD Matrices for Domain Adaptation
• A Second Order Non-Smooth Variational Model for Restoring Manifold-Valued Images
• Restoration of Manifold-Valued Images by Half-Quadratic Minimization
• Geometric Mean Metric Learning
• Geometry-aware Dynamic Movement Primitives
• Beyond Convexity -- Contraction and Global Convergence of Gradient Descent
• A Parallel Douglas Rachford Algorithm for Minimizing ROF-like Functionals on Images with Values in Symmetric Hadamard Manifolds
• Total Generalized Variation for Manifold-valued Data
• KKT Conditions, First-Order and Second-Order Optimization, and Distributed Optimization: Tutorial and Survey
• Learning Stable Robotic Skills on Riemannian Manifolds
• Fixed-point algorithms for learning determinantal point processes
• Bayesian Optimization Meets Riemannian Manifolds in Robot Learning
• Fast, asymptotically efficient, recursive estimation in a Riemannian manifold
• Separable Shape Tensors for Aerodynamic Design
• Momentum Improves Optimization on Riemannian Manifolds
• Interior-point methods on manifolds: theory and applications
• On Riemannian Optimization over Positive Definite Matrices with the Bures-Wasserstein Geometry
• Solving general elliptical mixture models through an approximate Wasserstein manifold
• Manifold Optimization for Gaussian Mixture Models
• Submodular Norms with Applications To Online Facility Location and Stochastic Probing
• Inference and Mixture Modeling with the Elliptical Gamma Distribution
• A Riemannian Newton Trust-Region Method for Fitting Gaussian Mixture Models
• Riemannian optimization on the simplex of positive definite matrices
• Deep Optimal Transport for Domain Adaptation on SPD Manifolds
• Efficient Hierarchical Bayesian Inference for Spatio-temporal Regression Models in Neuroimaging
• A Continuous-time Perspective for Modeling Acceleration in Riemannian Optimization
• Gradient Method for Optimization on Riemannian Manifolds with Lower Bounded Curvature
• Riemannian Stochastic Variance-Reduced Cubic Regularized Newton Method for Submanifold Optimization
• A Fast Graph Kernel Based Classification Method for Wireless Link Scheduling on Riemannian Manifold
• Geometric Machine Learning for Channel Covariance Estimation in Vehicular Networks
• Learning with symmetric positive definite matrices via generalized Bures-Wasserstein geometry
• Riemannian optimal identification method for linear systems with symmetric positive-definite matrix
• Spectral Flow on the Manifold of SPD Matrices for Multimodal Data Processing
• On the power means and Lawson-Lim means for positive invertible operators
• Riemannian information gradient methods for the parameter estimation of ECD: Some applications in image processing
• An inexact proximal point method for variational inequality on Hadamard manifolds
• Operator convex functions and their applications
• Vector Transport Free Riemannian LBFGS for Optimization on Symmetric Positive Definite Matrix Manifolds
• Manifold Optimization Assisted Gaussian Variational Approximation
• GeomNet: A Neural Network Based on Riemannian Geometries of SPD Matrix Space and Cholesky Space for 3D Skeleton-Based Interaction Recognition
• On the globalization of Riemannian Newton method
• Positive definite functions of noncommuting contractions, Hua-Bellman matrices, and a new distance metric