A note on computing matrix geometric means

A new definition is introduced for the matrix geometric mean of a set of k positive definite n×n matrices together with an iterative method for its computation. The iterative method is locally convergent with cubic convergence and requires O(n ³ k ²) arithmetic operations per step whereas the methods based on the symmetrization technique of Ando et al. (Linear Algebra Appl 385:305?C334, 2004) have complexity O(n ³ k!2 ^k ). The new mean is obtained from the properties of the centroid of a triangle rephrased in terms of geodesics in a suitable Riemannian geometry on the set of positive definite matrices. It satisfies most part of the ten properties stated by Ando, Li and Mathias; a counterexample shows that monotonicity is not fulfilled.     

Generalized Pólya–Szegö type inequalities for some non-commutative geometric means

In this paper, we shall show generalized Pólya–Szegö type inequalities of n positive invertible operators on a Hilbert space for any integer $n?3$ in terms of the following two typical non-commutative geometric means, that is, one is the higher order weighted geometric mean due to Lawson–Lim which is an extension of the Ando–Li–Mathias geometric mean, and the other is the weighted chaotic geometric mean. Among others, the Specht ratio plays an important role in our discussion, which is the upper bound of a ratio type reverse of the weighted arithmetic–geometric mean inequality.     

Some geometric properties of matrix means with respect to different metrics

In this paper we study the monotonicity, in-betweenness and in-sphere properties of matrix means with respect to Bures–Wasserstein, Hellinger and log-determinant metrics. More precisely, we show that the matrix power means (Kubo–Ando and non-Kubo–Ando extensions) satisfy the in-betweenness property in the Hellinger metric. We also show that for two positive definite matrices A and B, the curve of weighted Heron means, the geodesic curve of the arithmetic and the geometric mean lie inside the sphere centered at the geometric mean with the radius equal to half of the log-determinant distance between A and B.

     

Ando–Hiai type inequalities for multivariate operator means

ABSTRACT

We present several Ando–Hiai type inequalities for n-variable operator means for positive invertible operators. Ando–Hiai's inequalities given here are not only of the original type but also of the complementary type and of the reverse type involving the generalized Kantorovich constant.     

Multi-variable weighted geometric means of positive definite matrices

We define a family of weighted geometric means {G(t;ω;A)}_t∈[0,1]ⁿ where ω and A vary over all positive probability vectors in Rⁿ and n-tuples of positive definite matrices resp. Each of these weighted geometric means interpolates between the weighted ALM (t=0_n) and BMP (t=1_n) geometric means (ALM and BMP geometric means have been defined by Ando-Li-Mathias and Bini-Meini-Poloni, respectively.) We show that the weighted geometric means satisfy multidimensional versions of all properties that one would expect for a two-variable weighted geometric mean.     

Norm inequalities for matrix geometric means of positive definite matrices

We obtain eigenvalue inequalities for matrix geometric means of positive definite matrices. This implies matrix norm inequalities for unitarily invariant norms, which are considered as complementary to a series of norm inequalities among geometric means. We give complements of the Ando–Hiai type inequality for the Karcher mean by means of the generalized Kantorovich constant. Finally, we consider the monotonicity of the eigenvalue function for the Karcher mean.     

Improved cyclic reduction for solving queueing problems

The cyclic reduction technique (Buzbee et al., 1970), rephrased in functional form (Bini and Meini, 1996), provides a numerically stable, quadratically convergent method for solving the matrix equation X = ∑^{+ ∞} _i=0 X_i A_i, where the A_i's are nonnegative k × k matrices such that ∑^{+ ∞} _i=0 A_i is column stochastic. In this paper we propose a further improvement of the above method, based on a point-wise evaluation/interpolation at a suitable set of Fourier points, of the functional relations defining each step of cyclic reduction (Bini and Meini,1996). This new technique allows us to devise an algorithm based on FFT having a lower computational cost and a higher numerical stability. Numerical results and comparisons are provided. This revised version was published online in August 2006 with corrections to the Cover Date.     

Positive definiteness of Hermitian interval matrices

We present a new necessary and sufficient criterion to check the positive definiteness of Hermitian interval matrices. It is shown that an n×n Hermitian interval matrix is positive definite if and only if its 4^n-1(n-1)! specially chosen Hermitian vertex matrices are positive definite.     

Minimal condition number for positive definite Hankel matrices using semidefinite programming

We present a semidefinite programming approach for computing optimally conditioned positive definite Hankel matrices of order n. Unlike previous approaches, our method is guaranteed to find an optimally conditioned positive definite Hankel matrix within any desired tolerance. Since the condition number of such matrices grows exponentially with n, this is a very good test problem for checking the numerical accuracy of semidefinite programming solvers. Our tests show that semidefinite programming solvers using fixed double precision arithmetic are not able to solve problems with n>30. Moreover, the accuracy of the results for 24?n?30 is questionable. In order to accurately compute minimal condition number positive definite Hankel matrices of higher order, we use a Mathematica 6.0 implementation of the SDPHA solver that performs the numerical calculations in arbitrary precision arithmetic. By using this code, we have validated the results obtained by standard codes for n?24, and we have found optimally conditioned positive definite Hankel matrices up to n=100.     

Stopping criteria for the Ando-Li-Mathias and Bini-Meini-Poloni geometric means

We provide an upper bound for the number of iterations necessary to achieve a desired level of accuracy for the Ando-Li-Mathias [Linear Algebra Appl. 385 (2004) 305-334] and Bini-Meini-Poloni [Math. Comput. 79 (2010) 437-452] symmetrization procedures for computing the geometric mean of n positive definite matrices, where accuracy is measured by the spectral norm and the Thompson metric on the convex cone of positive definite matrices. It is shown that the upper bound for the number of iterations depends only on the diameter of the set of n matrices and the desired convergence tolerance. A striking result is that the upper bound decreases as n increases on any bounded region of positive definite matrices.     

Products of idempotent matrices

For some years it has been known that every singular square matrix over an arbitrary field F is a product of idempotent matrices over F. This paper quantifies that result to some extent. Main result: for every field F and every pair (n,k) of positive integers, an n×n matrix S over F is a product of k idempotent matrices over F iff rank(I ? S)?k· nullity S. The proof of the “if” part involves only elementary matrix operations and may thus be regarded as constructive. Corollary: (for every field F and every positive integer n) each singular n×n matrix over F is a product of n idempotent matrices over F, and there is a singular n×n matrix over F which is not a product of n ? 1 idempotent matrices.     

The resolvent average for positive semidefinite matrices

We define a new average - termed the resolvent average - for positive semidefinite matrices. For positive definite matrices, the resolvent average enjoys self-duality and it interpolates between the harmonic and the arithmetic averages, which it approaches when taking appropriate limits. We compare the resolvent average to the geometric mean. Some applications to matrix functions are also given.     

Nonstationary multisplittings with general weighting matrices for non-hermitian positive definite systems

We discuss the nonstationary multisplittings and two-stage multisplittings to solve the linear systems of algebraic equations Ax = b when the coefficient matrix is a non-Hermitian positive definite matrix, and establish the convergence theories with general weighting matrices. This not only eliminates the restrictive condition that it is usually assumed for scalar weighting matrices, but also generalizes it to a general positive definite matrix.     

Subtotally positive and Monge matrices

A real matrix is called k-subtotally positive if the determinants of all its submatrices of order at most k are positive. We show that for an m × n matrix, only mn inequalities determine such class for every k, 1 ? k ? min(m,n). Spectral properties of square k-subtotally positive matrices are studied. Finally, completion problems for 2-subtotally positive matrices and their additive counterpart, the anti-Monge matrices, are investigated. Since totally positive matrices are 2-subtotally positive as well, the presented necessary conditions for this completion problem are also necessary conditions for totally positive matrices.     

Finiteness of circulant weighing matrices of fixed weight

Let n be a fixed positive integer. Every circulant weighing matrix of weight n arises from what we call an irreducible orthogonal family of weight n. We show that the number of irreducible orthogonal families of weight n is finite and thus obtain a finite algorithm for classifying all circulant weighing matrices of weight n. We also show that, for every odd prime power q, there are at most finitely many proper circulant weighing matrices of weight q.     

Two inequalities involving Hadamard products of positive semi-definite Hermitian matrices

We extend two inequalities involving Hadamard products of positive definite Hermitian matrices to positive semi-definite Hermitian matrices. Simultaneously, we also show the sufficient conditions for equalities to hold. Moreover, some other matrix inequalities are also obtained. Our results and methods are different from those which are obtained by S. Liu in [J. Math. Anal. Appl. 243:458–463(2000)] and B.-Y. Wang et al. in [Lin. Alg. Appl. 302–303: 163–172(1999)].     

Hessenberg matrix for sums of Hermitian positive definite matrices and weighted shifts

In this work, we introduce an algebraic operation between bounded Hessenberg matrices and we analyze some of its properties. We call this operation m-sum and we obtain an expression for it that involves the Cholesky factorization of the corresponding Hermitian positive definite matrices associated with the Hessenberg components.This work extends a method to obtain the Hessenberg matrix of the sum of measures from the Hessenberg matrices of the individual measures, introduced recently by the authors for subnormal matrices, to matrices which are not necessarily subnormal.Moreover, we give some examples and we obtain the explicit formula for the m-sum of a weighted shift. In particular, we construct an interesting example: a subnormal Hessenberg matrix obtained as the m-sum of two not subnormal Hessenberg matrices.     

正定Hermite矩阵加权幂平均的行列式不等式

本文给出了m个正定Hermite矩阵加权幂平均的行列式的一个不等式，它是m个正数的加权益平均不等式的自然推广，也是正定Hermite矩阵行列式的凸性不等式的推广．