Information Geometry of Positive Definite Matrices

张士诚,孙华飞,李春辉

2009-01-01

Journal of Beijing Institute of Technology

Abstract:A new Riemannian metric for positive definite matrices is defined and its geomet ric structures are investigated by means of dual connections introduced to stati stical analysis by S. Amari. A few interesting results are obtained and some of those obtained by other authors are extended in our research.

What problem does this paper attempt to address?

Removable Singularity of Positive Mass Theorem with Continuous Metrics

Wenshuai Jiang,Weimin Sheng,Huaiyu Zhang

DOI: https://doi.org/10.1007/s00209-022-03081-w

2022-01-01

Mathematische Zeitschrift

Abstract:In this paper, we consider asymptotically flat Riemannian manifolds endowed with a continuous metric and the metric is smooth away from a compact subset with certain conditions. If the metric is Lipschitz and the scalar curvature is nonnegative away from a closed subset with (n - 1)-dimensional Hausdorff measure zero, then we prove that the ADM mass of each end is nonnegative. Furthermore, if the ADM mass of some end is zero, then we prove that the manifold is isometric to the Euclidean space. The Hausdorff measure condition is optimal, which confirms a conjecture of Lee in 2013. The proof of the nonnegativity is based on a smoothing argument with some new estimates such as a L-1-scalar curvature approximation estimates. The proof of the rigidity relies on the RCD theory where we show that the manifold has nonnegative Ricci curvature in RCD sense. The idea involving RCD theory in the proof turns out to be much more suitable in the study of the rigidity for continuous metrics. Furthermore, our result generalizes the result of Lee and LeFloch (Commun Math Phys 339(1):99-120, 2015) from spin to non-spin.
GEOMETRY MODULI SPACES

Kefeng Liu,Xiaofeng Sun,Shing-Tung Yau

2008-01-01

Abstract:In this paper we describe some recent results on the geometry of the moduli space of Riemann surfaces. We surveyed new and classical metrics on the moduli spaces of hyperbolic Riemann surfaces and their geometric properties. We then discussed the Mumford goodness and generalized goodness of various metrics on the moduli spaces and their deformation invariance. By combining with the dual Nakano negativity of the Weil-Petersson metric we derive various consequences such that the infinitesimal rigidity, the Gauss-Bonnet theorem and the log Chern number computations.
Positivity and vanishing theorems in complex and algebraic geometry

Kefeng Liu,Xiaokui Yang

2012-01-01

Abstract:In this thesis, we consider geometric properties of vector bundles arising from algebraic and Hermitian geometry. On vector bundles in algebraic geometry, such as ample, nef and globally generated vector bundles, we are able to construct positive Hermitian metrics in different senses(e.g. Griffiths-positive, Nakano-positive and dual-Nakano-positive) by L2-method and deduce many new vanishing theorems for them by analytic method instead of the Le Potier-Leray spectral sequence method. On Hermitian manifolds, we find that the second Ricci curvature tensors of various metric connections are closely related to the geometry of Hermitian manifolds. We can derive various vanishing theorems for Hermitian manifolds and also for complex vector bundles over Hermitian manifolds by their second Ricci curvature tensors. We also introduce a natural geometric flow on Hermitian manifolds by using the second Ricci curvature tensor.
New results on the geometry of the moduli space of Riemann surfaces

KeFeng Liu,XiaoFeng Sun,Shing-Tung Yau

DOI: https://doi.org/10.1007/s11425-008-0063-6

2008-01-01

Abstract:We briefly survey our recent results about the Mumford goodness of several canonical metrics on the moduli spaces of Riemann surfaces, including the Weil-Petersson metric, the Ricci metric, the Perturbed Ricci metric and the Kahler-Einstein metric. We prove the dual Nakano negativity of the Weil-Petersson metric. As applications of these results we deduce certain important results about the L 2 -cohomology groups of the logarithmic tangent bundle over the compactified moduli spaces.
Geometry of Hermitian Manifolds

Ke-Feng Liu,Xiao-Kui Yang

DOI: https://doi.org/10.1142/s0129167x12500553

2011-01-01

International Journal of Mathematics

Abstract:On Hermitian manifolds, the second Ricci curvature tensors of various metric connections are closely related to the geometry of Hermitian manifolds. By refining the Bochner formulas for any Hermitian complex vector bundle (and Riemannian real vector bundle) with an arbitrary metric connection over a compact Hermitian manifold, we can derive various vanishing theorems for Hermitian manifolds and complex vector bundles by the second Ricci curvature tensors. We will also introduce a natural geometric flow on Hermitian manifolds by using the second Ricci curvature tensor.
Riemannian Gaussian distributions on the space of positive-definite quaternion matrices

Salem Said,Nicolas Le Bihan,Jonathan H. Manton

DOI: https://doi.org/10.48550/arXiv.1703.09940

2017-03-29

Abstract:Recently, Riemannian Gaussian distributions were defined on spaces of positive-definite real and complex matrices. The present paper extends this definition to the space of positive-definite quaternion matrices. In order to do so, it develops the Riemannian geometry of the space of positive-definite quaternion matrices, which is shown to be a Riemannian symmetric space of non-positive curvature. The paper gives original formulae for the Riemannian metric of this space, its geodesics, and distance function. Then, it develops the theory of Riemannian Gaussian distributions, including the exact expression of their probability density, their sampling algorithm and statistical inference.

Statistics Theory
Transversely Hessian foliations and information geometry

Michel Nguiffo Boyom,Robert A. Wolak

DOI: https://doi.org/10.1142/s0129167x16500920

2016-10-01

International Journal of Mathematics

Abstract:A family of probability distributions parametrized by an open domain [Formula: see text] in [Formula: see text] defines the Fisher information matrix on this domain which is positive semi-definite. In information geometry, the standard assumption has been that the Fisher information matrix tensor is positive definite defining in this way a Riemannian metric on [Formula: see text]. It seems to be quite a strong condition. In general, not much can be said about the Fisher information matrix tensor. To develop a more general theory, we weaken the assumption and replace “positive definite” by the existence of a suitable torsion-free connection. It permits us to define naturally a foliation with a transversely Hessian structure. We develop the theory of transversely Hessian foliations along the lines of the classical theory.

mathematics
Riemannian Metric Learning for Symmetric Positive Definite Matrices

Raviteja Vemulapalli,David W. Jacobs

DOI: https://doi.org/10.48550/arXiv.1501.02393

2015-01-11

Abstract:Over the past few years, symmetric positive definite (SPD) matrices have been receiving considerable attention from computer vision community. Though various distance measures have been proposed in the past for comparing SPD matrices, the two most widely-used measures are affine-invariant distance and log-Euclidean distance. This is because these two measures are true geodesic distances induced by Riemannian geometry. In this work, we focus on the log-Euclidean Riemannian geometry and propose a data-driven approach for learning Riemannian metrics/geodesic distances for SPD matrices. We show that the geodesic distance learned using the proposed approach performs better than various existing distance measures when evaluated on face matching and clustering tasks.

Computer Vision and Pattern Recognition,Machine Learning
Positive geometries and canonical forms

Nima Arkani-Hamed,Yuntao Bai,Thomas Lam

DOI: https://doi.org/10.1007/jhep11(2017)039

IF: 6.379

2017-11-01

Journal of High Energy Physics

Abstract:A bstract Recent years have seen a surprising connection between the physics of scattering amplitudes and a class of mathematical objects — the positive Grassmannian, positive loop Grassmannians, tree and loop Amplituhedra — which have been loosely referred to as “positive geometries”. The connection between the geometry and physics is provided by a unique differential form canonically determined by the property of having logarithmic singularities (only) on all the boundaries of the space, with residues on each boundary given by the canonical form on that boundary. The structures seen in the physical setting of the Amplituhedron are both rigid and rich enough to motivate an investigation of the notions of “positive geometries” and their associated “canonical forms” as objects of study in their own right, in a more general mathematical setting. In this paper we take the first steps in this direction. We begin by giving a precise definition of positive geometries and canonical forms, and introduce two general methods for finding forms for more complicated positive geometries from simpler ones — via “triangulation” on the one hand, and “push-forward” maps between geometries on the other. We present numerous examples of positive geometries in projective spaces, Grassmannians, and toric, cluster and flag varieties, both for the simplest “simplex-like” geometries and the richer “polytope-like” ones. We also illustrate a number of strategies for computing canonical forms for large classes of positive geometries, ranging from a direct determination exploiting knowledge of zeros and poles, to the use of the general triangulation and push-forward methods, to the representation of the form as volume integrals over dual geometries and contour integrals over auxiliary spaces. These methods yield interesting representations for the canonical forms of wide classes of positive geometries, ranging from the simplest Amplituhedra to new expressions for the volume of arbitrary convex polytopes.

physics, particles & fields
Cartan-Schouten metrics for information geometry and machine learning

Andre Diatta,Bakary Manga,Fatimata Sy

2024-08-28

Abstract:We study Cartan-Schouten metrics, explore invariant dual connections, and propose them as models for Information Geometry. Based on the underlying Riemannian barycenter and the biinvariant mean of Lie groups, we subsequently propose a new parametric mean for data science and machine learning which comes with several advantages compared to traditional tools such as the arithmetic mean, median, mode, expectation, least square method, maximum likelihood, linear regression. We call a metric on a Lie group, a Cartan-Schouten metric, if its Levi-Civita connection is biinvariant, so every 1-parameter subgroup through the unit is a geodesic. Except for not being left or right invariant in general, Cartan-Schouten metrics enjoy the same geometry as biinvariant metrics, since they share the same Levi-Civita connection. To bypass the non-invariance apparent drawback, we show that Cartan-Schouten metrics are completely determined by their value at the unit. We give an explicit formula for recovering them from their value at the unit, thus making them much less computationally demanding, compared to general metrics on manifolds. Furthermore, Lie groups with Cartan-Schouten metrics are complete Riemannian or pseudo-Riemannian manifolds. We give a complete characterization of Lie groups with Riemannian or Lorentzian Cartan-Schouten metrics. Cartan-Schouten metrics are in abundance on 2-nilpotent Lie groups. Namely, on every 2-nilpotent Lie group, there is a 1-1 correspondence between the set of left invariant metrics and that of Cartan-Schouten metrics.

Differential Geometry,Information Theory,Mathematical Physics,Probability,Statistics Theory
Pulsar Parallaxes at 5 GHz with the Very Long Baseline Array

S. Chatterjee,J. M. Cordes,W. H. T. Vlemmings,Z. Arzoumanian,W. M. Goss,T. J. W. Lazio

DOI: https://doi.org/10.1086/381748

2003-12-02

Abstract:We present the first pulsar parallaxes measured with phase-referenced pulsar VLBI observations at 5 GHz. Due to the steep spectra of pulsars, previous astrometric measurements have been at lower frequencies. However, the strongest pulsars can be observed at 5 GHz, offering the benefit of lower combined ionospheric and tropospheric phase errors, which usually limit VLBI astrometric accuracy. The pulsars B0329+54, B0355+54 and B1929+10 were observed for 7 epochs spread evenly over 2 years. For B0329+54, large systematic errors lead to only an upper limit on the parallax (pi < 1.5 mas). A new proper motion and parallax were measured for B0355+54 (pi = 0.91 +- 0.16 mas), implying a distance of 1.04+0.21-0.16 kpc and a transverse velocity of 61+12-9 km/s. The parallax and proper motion for B1929+10 were significantly improved (pi = 2.77 +- 0.07 mas), yielding a distance of 361+10-8 pc and a transverse velocity of 177+4-5 km/s. We demonstrate that the astrometric errors are correlated with the angular separation between the phase reference calibrator and the target source, with significantly lower errors at 5 GHz compared to 1.6 GHz. Finally, based on our new distance determinations for B1929+10 and B0355+54, we derive or constrain the luminosities of each pulsar at high energies. We show that, for thermal emission models, the emitting area for X-rays from PSR B1929+10 is roughly consistent with the canonical size for a heated polar cap, and that the conversion of spin-down power to gamma-ray luminosity in B0355+54 must be low. The new proper motion for B1929+10 also implies that its progenitor is unlikely to have been the binary companion of the runaway O-star zeta-Ophiuchi.

Astrophysics
Positive positive-definite functions and measures on locally compact abelian groups

Alexandr Borisov

DOI: https://doi.org/10.48550/arXiv.math/9906126

1999-06-19

Abstract:The author was recently able to provide a cohomological interpretation of Tate's Riemann-Roch formula for number fields using some new harmonic analysis objects, ghost-spaces. When trying to investigate these objects in general, we realized the importance of functions and measures on locally compact abelian groups that are both positive and positive-definite at the same time. It looks like this class of functions and measures was not systematically studied before. The goal of this paper is to partially fill in this gap. We answer some of the natural questions involving these functions and measures, especially those that satisfy some extra integrability conditions. We also study some operations and constructions involving these functions and measures. There are several very interesting open questions, that we are only able to point out at this moment. In particular, the structure of the cone of such functions is not clear even when the group is just $\R.$

Functional Analysis
Transport information geometry I: Riemannian calculus on probability simplex

Wuchen Li

DOI: https://doi.org/10.48550/arXiv.1803.06360

2020-04-09

Abstract:We formulate the Riemannian calculus of the probability set embedded with $L^2$-Wasserstein metric. This is an initial work of transport information geometry. Our investigation starts with the probability simplex (probability manifold) supported on vertices of a finite graph. The main idea is to embed the probability manifold as a submanifold of the positive measure space with a nonlinear metric tensor. Here the nonlinearity comes from the linear weighted Laplacian operator. By this viewpoint, we establish torsion-free Christoffel symbols, Levi-Civita connections, curvature tensors and volume forms in the probability manifold by Euclidean coordinates. As a consequence, the Jacobi equation, Laplace-Beltrami and Hessian operators on the probability manifold are derived. These geometric computations are also provided in the infinite-dimensional density space (density manifold) supported on a finite-dimensional manifold. In particular, an identity is given connecting the Baker-{É}mery $\Gamma_2$ operator (carr{é} du champ it{é}r{é}) by connecting Fisher-Rao information metric and optimal transport metric. Several examples are demonstrated.

Differential Geometry,Information Theory
Positive definite matrices and the S-divergence

Suvrit Sra

DOI: https://doi.org/10.1090/proc/12953

IF: 0.9709

2015-10-22

Proceedings of the American Mathematical Society

Abstract:Hermitian positive definite (hpd) matrices form a self-dual convex cone whose interior is a Riemannian manifold of nonpositive curvature. The manifold view comes with a natural distance function but the conic view does not. Thus, drawing motivation from convex optimization we introduce the S-divergence , a distance-like function on the cone of hpd matrices. We study basic properties of the S-divergence and explore its connections to the Riemannian distance. In particular, we show that (i) its square-root is a distance, and (ii) it exhibits numerous nonpositive-curvature-like properties.

mathematics, applied
Positivity of Ma-Trudinger-Wang Curvature on Riemannian Surfaces

Shi-Zhong Du,Qi-Rui Li

DOI: https://doi.org/10.1007/s00526-013-0684-7

2013-01-01

Calculus of Variations and Partial Differential Equations

Abstract:In this paper we consider the optimal transportation on Riemannian surfaces when the cost function is squared distance. The main ingredient is the verification of MTW condition. It is known that MTW condition holds if Gauss curvature is sufficiently close to $1$ in $C^{2}$ norm. In this paper we give an explicit condition on Gauss curvature such that MTW condition is satisfied.
On locally symmetric polynomial metrics: Riemannian and Finslerian surfaces

Csaba Vincze,Márk Oláh,Ábris Nagy

2024-03-15

Abstract:In the paper we investigate locally symmetric polynomial metrics in special cases of Riemannian and Finslerian surfaces. The Riemannian case will be presented by a collection of basic results (regularity of second root metrics) and formulas up to Gauss curvature. In case of Finslerian surfaces we formulate necessary and sufficient conditions for a locally symmetric fourth root metric in 2D to be positive definite. They are given in terms of the coefficients of the polynomial metric to make checking the positive definiteness as simple and direct as possible. Explicit examples are also presented. The situation is more complicated in case of spaces of dimension more than two. Some necessary conditions and an explicit example are given for a positive definite locally symmetric polynomial metric in 3D. Computations are supported by the MAPLE mathematics software (LinearAlgebra).

Differential Geometry
Information geometry of diffeomorphism groups

Boris Khesin,Gerard Misiołek,Klas Modin

2024-11-06

Abstract:The study of diffeomorphism groups and their applications to problems in analysis and geometry has a long history. In geometric hydrodynamics, pioneered by V.~Arnold in the 1960s, one considers an ideal fluid flow as the geodesic motion on the infinite-dimensional group of volume-preserving diffeomorphisms of the fluid domain with respect to the metric defined by the kinetic energy. Similar considerations on the space of densities lead to a geometric description of optimal mass transport and the Kantorovich-Wasserstein metric. Likewise, information geometry associated with the Fisher-Rao metric and the Hellinger distance has an equally beautiful infinite-dimensional geometric description and can be regarded as a higher-order Sobolev analogue of optimal transportation. In this work we review various metrics on diffeomorphism groups relevant to this approach and introduce appropriate topology, smooth structures and dynamics on the corresponding infinite-dimensional manifolds. Our main goal is to demonstrate how, alongside topological hydrodynamics, Hamiltonian dynamics and optimal mass transport, information geometry with its elaborate toolbox has become yet another exciting field for applications of geometric analysis on diffeomorphism groups.

Differential Geometry,Information Theory,Mathematical Physics
Geometric mean for T-positive definite tensors and associated Riemannian geometry

Jeong-Hoon Ju,Taehyeong Kim,Yeongrak Kim,Hayoung Choi

2024-03-30

Abstract:In this paper, we generalize the geometric mean of two positive definite matrices to that of third-order tensors using the notion of T-product. Specifically, we define the geometric mean of two T-positive definite tensors and verify several properties that "mean" should satisfy including the idempotence and the commutative property, and so on. Moreover, it is shown that the geometric mean is a unique T-positive definite solution of an algebraic Riccati tensor equation and can be expressed as solutions of algebraic Riccati matrix equations. In addition, we investigate the Riemannian manifold associated with the geometric mean for T-positive definite tensors, considering it as a totally geodesic embedded submanifold of the Riemannian manifold associated with the case of matrices. It is particularly shown that the geometric mean of two T-positive definite tensors is the midpoint of a unique geodesic joining the tensors, and the manifold is a Cartan-Hadamard-Riemannian manifold.

Numerical Analysis,Differential Geometry,Functional Analysis
Information Geometry for the Working Information Theorist

Kumar Vijay Mishra,M. Ashok Kumar,Ting-Kam Leonard Wong

2023-10-06

Abstract:Information geometry is a study of statistical manifolds, that is, spaces of probability distributions from a geometric perspective. Its classical information-theoretic applications relate to statistical concepts such as Fisher information, sufficient statistics, and efficient estimators. Today, information geometry has emerged as an interdisciplinary field that finds applications in diverse areas such as radar sensing, array signal processing, quantum physics, deep learning, and optimal transport. This article presents an overview of essential information geometry to initiate an information theorist, who may be unfamiliar with this exciting area of research. We explain the concepts of divergences on statistical manifolds, generalized notions of distances, orthogonality, and geodesics, thereby paving the way for concrete applications and novel theoretical investigations. We also highlight some recent information-geometric developments, which are of interest to the broader information theory community.

Information Theory,Machine Learning,Signal Processing,Differential Geometry
On Certain Positive Semidefinite Matrices of Special Functions

Ruiming Zhang

DOI: https://doi.org/10.48550/arXiv.1603.05512

2016-03-20

Abstract:Special functions are often defined as a Fourier or Laplace transform of a positive measure, and the positivity of the measure manifests as positive definiteness of certain matrices. The purpose of this expository note is to give a sample of such positive definite matrices to demonstrate this connection for some well-known special functions such as Gamma, Beta, hypergeometric, theta, elliptic, zeta and basic hypergeometric functions.

Classical Analysis and ODEs

Information Geometry of Positive Definite Matrices

Removable Singularity of Positive Mass Theorem with Continuous Metrics

GEOMETRY MODULI SPACES

Positivity and vanishing theorems in complex and algebraic geometry

New results on the geometry of the moduli space of Riemann surfaces

Geometry of Hermitian Manifolds

Riemannian Gaussian distributions on the space of positive-definite quaternion matrices

Transversely Hessian foliations and information geometry

Riemannian Metric Learning for Symmetric Positive Definite Matrices

Positive geometries and canonical forms

Cartan-Schouten metrics for information geometry and machine learning

Pulsar Parallaxes at 5 GHz with the Very Long Baseline Array

Positive positive-definite functions and measures on locally compact abelian groups

Transport information geometry I: Riemannian calculus on probability simplex

Positive definite matrices and the S-divergence

Positivity of Ma-Trudinger-Wang Curvature on Riemannian Surfaces

On locally symmetric polynomial metrics: Riemannian and Finslerian surfaces

Information geometry of diffeomorphism groups

Geometric mean for T-positive definite tensors and associated Riemannian geometry

Information Geometry for the Working Information Theorist

On Certain Positive Semidefinite Matrices of Special Functions