Multivariate Gaussians, semidefinite matrix completion, and convex algebraic geometry

DOI10.1007/S10463-010-0295-4zbMATH Open1440.62255arXiv0906.3529OpenAlexW2040714878MaRDI QIDQ2634801FDOQ2634801

Authors: Bernd Sturmfels, Caroline Uhler 

Publication date: 19 February 2016

Published in: Annals of the Institute of Statistical Mathematics (Search for Journal in Brave)

Abstract: We study multivariate normal models that are described by linear constraints on the inverse of the covariance matrix. Maximum likelihood estimation for such models leads to the problem of maximizing the determinant function over a spectrahedron, and to the problem of characterizing the image of the positive definite cone under an arbitrary linear projection. These problems at the interface of statistics and optimization are here examined from the perspective of convex algebraic geometry.

Full work available at URL: https://arxiv.org/abs/0906.3529

Recommendations

• The maximum likelihood degree of linear spaces of symmetric matrices
• Maximum likelihood degree of the two-dimensional linear Gaussian covariance model
• Maximum likelihood for matrices with rank constraints
• Maximum likelihood estimation for linear Gaussian covariance models
• The maximum likelihood degree

zbMATH Keywords

multivariate normal distributionmaximum likelihood estimationconvex algebraic geometrysemidefinite matrix completiondual convex conedual projective variety

Mathematics Subject Classification ID

Multivariate analysis (62H99) Graphs and linear algebra (matrices, eigenvalues, etc.) (05C50) Estimation in multivariate analysis (62H12) Algebraic statistics (62R01) Positive matrices and their generalizations; cones of matrices (15B48) Gröbner bases; other bases for ideals and modules (e.g., Janet and border bases) (13P10)

Cites Work

• SINGULAR
• Title not available (Why is that?)
• Covariance hypothesis which are linear in both the covariance and the inverse covariance
• Title not available (Why is that?)
• Algebraic Statistics for Computational Biology
• Lectures on algebraic statistics
• Trek separation for Gaussian graphical models
• A Combinatorial Model for Series-Parallel Networks
• Positive definite completions of partial Hermitian matrices
• The algebraic degree of semidefinite programming
• Ideals with sliding depth
• Positive semidefinite matrices with a given sparsity pattern
• The real positive definite completion problem for a simple cycle
• On the sparsity order of a graph and its deficiency in chordality
• Facing up to arrangements: face-count formulas for partitions of space by hyperplanes
• The real positive definite completion problem: cycle completability
• A broken circuit ring
• Title not available (Why is that?)
• Algebraic geometry of Bayesian networks
• Title not available (Why is that?)
• Determinant Maximization with Linear Matrix Inequality Constraints
• Algebras generated by reciprocals of linear forms
• Polyhedral conditions for the nonexistence of the MLE for hierarchical log-linear models
• Title not available (Why is that?)
• Determinantal ideals of linear type of a generic symmetric matrix
• Algebraic geometry of Gaussian Bayesian networks
• On quasi-complete intersections
• Hilbert series of residual intersections

Cited In (42)

• Title not available (Why is that?)
• The tropical critical points of an affine matroid
• Complete quadrics: Schubert calculus for Gaussian models and semidefinite programming
• Equations and multidegrees for inverse symmetric matrix pairs
• On a parametrization of positive semidefinite matrices with zeros
• ORBITOPES
• Estimating linear covariance models with numerical nonlinear algebra
• Discriminant analysis with Gaussian graphical tree models
• Title not available (Why is that?)
• Discriminants and nonnegative polynomials
• Symmetrically Colored Gaussian Graphical Models with Toric Vanishing Ideals
• Algebraic boundaries of convex semi-algebraic sets
• Exponential varieties
• Differential equations for Gaussian statistical models with rational maximum likelihood estimator
• Directed Gaussian graphical models with toric vanishing ideals
• Real rank geometry of ternary forms
• The central curve in linear programming
• Pencils of Quadrics: Old and New
• Generic Spectrahedral Shadows
• Extremal positive semidefinite matrices whose sparsity pattern is given by graphs without \(K_{5}\) minors
• Logarithmic Voronoi cells for Gaussian models
• Nonlinear algebra and applications
• Multidegrees, prime ideals, and non-standard gradings
• Real algebraic geometry with a view toward Koopman operator methods. Abstracts from the workshop held March 12--17, 2023
• Gaussian graphical models with toric vanishing ideals
• Bad projections of the PSD cone
• Symmetries in directed Gaussian graphical models
• Ideals of submaximal minors of sparse symmetric matrices
• Inverting catalecticants of ternary quartics
• Reciprocal Maximum Likelihood Degrees of Brownian Motion Tree Models
• The degree of the central curve in semidefinite, linear, and quadratic programming
• Maximum Likelihood Degree, Complete Quadrics, and $\mathbb{C}^*$-Action
• Title not available (Why is that?)
• Coloured Graphical Models and their Symmetries
• Initial values of ML-degree polynomials
• A review of Gaussian Markov models for conditional independence
• The leading coefficient of Lascoux polynomials
• Gibbs manifolds
• Gaussian Likelihood Geometry of Projective Varieties
• Title not available (Why is that?)
• Lefschetz properties in algebra, geometry and combinatorics. Abstracts from the workshop held September 27 -- October 3, 2020 (hybrid meeting)
• Geometry of maximum likelihood estimation in Gaussian graphical models

Uses Software

• Macaulay2
• SINGULAR
• CVX
• gRc
• surfex

This page was built for publication: Multivariate Gaussians, semidefinite matrix completion, and convex algebraic geometry

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2634801)