Extremal eigenvalues and eigenvectors of deformed Wigner matrices

We consider random matrices of the form \(H = W + \lambda V, \lambda \in {\mathbb {R}}^+\), where \(W\) is a real symmetric or complex Hermitian Wigner matrix of size \(N\) and \(V\) is a real bounded diagonal random matrix of size \(N\) with i.i.d. entries that are independent of \(W\). We assume subexponential decay of the distribution of the matrix entries of \(W\) and we choose \(\lambda \sim 1\), so that the eigenvalues of \(W\) and \(\lambda V\) are typically of the same order. Further, we assume that the density of the entries of \(V\) is supported on a single interval and is convex near the edges of its support. In this paper we prove that there is \(\lambda _+\in {\mathbb {R}}^+\) such that the largest eigenvalues of \(H\) are in the limit of large \(N\) determined by the order statistics of \(V\) for \(\lambda >\lambda _+\). In particular, the largest eigenvalue of \(H\) has a Weibull distribution in the limit \(N\rightarrow \infty \) if \(\lambda >\lambda _+\). Moreover, for \(N\) sufficiently large, we show that the eigenvectors associated to the largest eigenvalues are partially localized for \(\lambda >\lambda _+\), while they are completely delocalized for \(\lambda <\lambda _+\). Similar results hold for the lowest eigenvalues.     

Algorithms for positive semidefinite factorization

This paper considers the problem of positive semidefinite factorization (PSD factorization), a generalization of exact nonnegative matrix factorization. Given an m-by-n nonnegative matrix X and an integer k, the PSD factorization problem consists in finding, if possible, symmetric k-by-k positive semidefinite matrices \(\{A^1,\ldots ,A^m\}\) and \(\{B^1,\ldots ,B^n\}\) such that \(X_{i,j}=\text {trace}(A^iB^j)\) for \(i=1,\ldots ,m\), and \(j=1,\ldots ,n\). PSD factorization is NP-hard. In this work, we introduce several local optimization schemes to tackle this problem: a fast projected gradient method and two algorithms based on the coordinate descent framework. The main application of PSD factorization is the computation of semidefinite extensions, that is, the representations of polyhedrons as projections of spectrahedra, for which the matrix to be factorized is the slack matrix of the polyhedron. We compare the performance of our algorithms on this class of problems. In particular, we compute the PSD extensions of size \(k=1+ \lceil \log _2(n) \rceil \) for the regular n-gons when \(n=5\), 8 and 10. We also show how to generalize our algorithms to compute the square root rank (which is the size of the factors in a PSD factorization where all factor matrices \(A^i\) and \(B^j\) have rank one) and completely PSD factorizations (which is the special case where the input matrix is symmetric and equality \(A^i=B^i\) is required for all i).     

Hyperinvariant Subspace Problem for Some Classes of Operators

An n-normal operator may be defined as an \(n \times n\) operator matrix with entries that are mutually commuting normal operators and an operator \(T \in \mathcal {B(H)}\) is quasi-nM-hyponormal (for \(n \in \mathbb {N}\)) if it is unitarily equivalent to an \(n \times n\) upper triangular operator matrix \((T_{ij})\) acting on \(\mathcal {K}^{(n)}\), where \(\mathcal {K}\) is a separable complex Hilbert space and the diagonal entries \(T_{jj}\) \((j = 1,2,\ldots , n)\) are M-hyponormal operators in \(\mathcal {B(K)}\). This is an extended notion of n-normal operators. We prove a necessary and sufficient condition for an \(n \times n\) triangular operator matrix to have Bishop’s property \((\beta )\). This leads us to study the hyperinvariant subspace problem for an \(n \times n\) triangular operator matrix.     

Partial regularity and smooth topology-preserving approximations of rough domains

For a bounded domain \(\Omega \subset {\mathbb R}^m, m\ge 2,\) of class \(C^0\), the properties are studied of fields of ‘good directions’, that is the directions with respect to which \(\partial \Omega \) can be locally represented as the graph of a continuous function. For any such domain there is a canonical smooth field of good directions defined in a suitable neighbourhood of \(\partial \Omega \), in terms of which a corresponding flow can be defined. Using this flow it is shown that \(\Omega \) can be approximated from the inside and the outside by diffeomorphic domains of class \(C^\infty \). Whether or not the image of a general continuous field of good directions (pseudonormals) defined on \(\partial \Omega \) is the whole of \(S^{m-1}\) is shown to depend on the topology of \(\Omega \). These considerations are used to prove that if \(m=2,3\), or if \(\Omega \) has nonzero Euler characteristic, there is a point \(P\in \partial \Omega \) in the neighbourhood of which \(\partial \Omega \) is Lipschitz. The results provide new information even for more regular domains, with Lipschitz or smooth boundaries.     

The fixed points of the multivariate smoothing transform

Let \((\mathbf {T}_1, \mathbf {T}_2, \ldots )\) be a sequence of random \(d\times d\) matrices with nonnegative entries, and let Q be a random vector with nonnegative entries. Consider random vectors \(X\) with nonnegative entries, satisfying

$$\begin{aligned} X\mathop {=}\limits ^{{\mathcal {L}}}\sum _{i \ge 1} \mathbf {T}_i X_i + Q, \end{aligned}$$

(*)

where \(\mathop {=}\limits ^{{\mathcal {L}}}\) denotes equality of the corresponding laws, \((X_i)_{i \ge 1}\) are i.i.d. copies of \(X\) and independent of \((Q, \mathbf {T}_1, \mathbf {T}_2, \ldots )\). For \(d=1\), this equation, known as fixed point equation of the smoothing transform, has been intensively studied. Under assumptions similar to the one-dimensional case, we obtain a complete characterization of all solutions \(X\) to (*) in the non-critical case, and existence results in the critical case.

     

A generalization of zero divisor graphs associated to commutative rings

Let R be a commutative ring with a nonzero identity element. For a natural number n, we associate a simple graph, denoted by \(\Gamma ^n_R\), with \(R^n\backslash \{0\}\) as the vertex set and two distinct vertices X and Y in \(R^n\) being adjacent if and only if there exists an \(n\times n\) lower triangular matrix A over R whose entries on the main diagonal are nonzero and one of the entries on the main diagonal is regular such that \(X^TAY=0\) or \(Y^TAX=0\), where, for a matrix \(B, B^T\) is the matrix transpose of B. If \(n=1\), then \(\Gamma ^n_R\) is isomorphic to the zero divisor graph \(\Gamma (R)\), and so \(\Gamma ^n_R\) is a generalization of \(\Gamma (R)\) which is called a generalized zero divisor graph of R. In this paper, we study some basic properties of \(\Gamma ^n_ R\). We also determine all isomorphic classes of finite commutative rings whose generalized zero divisor graphs have genus at most three.     

An <Emphasis Type="Italic">A</Emphasis>-invariant subspace for bipartite distance-regular graphs with exactly two irreducible <Emphasis Type="Italic">T</Emphasis>-modules with endpoint 2, both thin

Let \(\Gamma \) denote a bipartite distance-regular graph with vertex set X, diameter \(D \ge 4\), and valency \(k \ge 3\). Let \({{\mathbb {C}}}^X\) denote the vector space over \({{\mathbb {C}}}\) consisting of column vectors with entries in \({{\mathbb {C}}}\) and rows indexed by X. For \(z \in X\), let \({{\widehat{z}}}\) denote the vector in \({{\mathbb {C}}}^X\) with a 1 in the z-coordinate, and 0 in all other coordinates. Fix a vertex x of \(\Gamma \) and let \(T = T(x)\) denote the corresponding Terwilliger algebra. Assume that up to isomorphism there exist exactly two irreducible T-modules with endpoint 2, and they both are thin. Fix \(y \in X\) such that \(\partial (x,y)=2\), where \(\partial \) denotes path-length distance. For \(0 \le i,j \le D\) define \(w_{ij}=\sum {{\widehat{z}}}\), where the sum is over all \(z \in X\) such that \(\partial (x,z)=i\) and \(\partial (y,z)=j\). We define \(W=\mathrm{span}\{w_{ij} \mid 0 \le i,j \le D\}\). In this paper we consider the space \(MW=\mathrm{span}\{mw \mid m \in M, w \in W\}\), where M is the Bose–Mesner algebra of \(\Gamma \). We observe that MW is the minimal A-invariant subspace of \({{\mathbb {C}}}^X\) which contains W, where A is the adjacency matrix of \(\Gamma \). We show that \(4D-6 \le \mathrm{dim}(MW) \le 4D-2\). We display a basis for MW for each of these five cases, and we give the action of A on these bases.     

Simultaneous power factorization in modules over Banach algebras

Let A be a Banach algebra with a bounded left approximate identity \(\{e_\lambda \}_{\lambda \in \Lambda }\), let \(\pi \) be a continuous representation of A on a Banach space X, and let S be a non-empty subset of X such that \(\lim _{\lambda }\pi (e_\lambda )s=s\) uniformly on S. If S is bounded, or if \(\{e_\lambda \}_{\lambda \in \Lambda }\) is commutative, then we show that there exist \(a\in A\) and maps \(x_n: S\rightarrow X\) for \(n\ge 1\) such that \(s=\pi (a^n)x_n(s)\) for all \(n\ge 1\) and \(s\in S\). The properties of \(a\in A\) and the maps \(x_n\), as produced by the constructive proof, are studied in some detail. The results generalize previous simultaneous factorization theorems as well as Allan and Sinclair’s power factorization theorem. In an ordered context, we also consider the existence of a positive factorization for a subset of the positive cone of an ordered Banach space that is a positive module over an ordered Banach algebra with a positive bounded left approximate identity. Such factorizations are not always possible. In certain cases, including those for positive modules over ordered Banach algebras of bounded functions, such positive factorizations exist, but the general picture is still unclear. Furthermore, simultaneous pointwise power factorizations for sets of bounded maps with values in a Banach module (such as sets of bounded convergent nets) are obtained. A worked example for the left regular representation of \(\mathrm {C}_0({\mathbb R})\) and unbounded S is included.     

Matrix Gegenbauer Polynomials: The $$2\times 2$$ Fundamental Cases

In this paper, we exhibit explicitly a sequence of \(2 \times 2\) matrix valued orthogonal polynomials with respect to a weight \(W_{p,n}\), for any pair of real numbers p and n such that \(0<p<n\). The entries of these polynomiales are expressed in terms of the Gegenbauer polynomials \(C_k^\lambda \). The corresponding three-term recursion relations are also given, and we make some studies of the algebra of differential operators associated with the weight \(W_{p,n}\).     

$$u\tau $$-convergence in locally solid vector lattices

Let \((x_\alpha )\) be a net in a locally solid vector lattice \((X,\tau )\); we say that \((x_\alpha )\) is unbounded \(\tau \)-convergent to a vector \(x\in X\) if \(|x_\alpha -x |\wedge w \xrightarrow {\tau } 0\) for all \(w\in X_+\). In this paper, we study general properties of unbounded \(\tau \)-convergence (shortly \(u\tau \)-convergence). \(u\tau \)-convergence generalizes unbounded norm convergence and unbounded absolute weak convergence in normed lattices that have been investigated recently. We introduce \(u\tau \)-topology and briefly study metrizability and completeness of this topology.     

On the quasi-arithmetic Gauss-type iteration

For a sequence of continuous, monotone functions \(f_1,\ldots ,f_n :I \rightarrow \mathbb {R}\) (I is an interval) we define the mapping \(M :I^n \rightarrow I^n\) as a Cartesian product of quasi-arithmetic means generated by \(f_j\)-s. It is known that, for every initial vector, the iteration sequence of this mapping tends to the diagonal of \(I^n\). We will prove that whenever all \(f_j\)-s are \(\mathcal {C}^2\) with nowhere vanishing first derivative, then this convergence is quadratic. Furthermore, the limit \(\frac{\text {Var}\, M^{k+1}(v)}{(\text {Var}\, M^{k}(v))^2}\) will be calculated in a nondegenerated case.     

Standard Jordan partitions with three parameters

We extend previous work on standard two-parameter Jordan partitions by Barry (Commun Algebra 43:4231–4246, 2015) to three parameters. Let \(J_r\) denote an \(r \times r\) matrix with minimal polynomial \((t-1)^r\) over a field F of characteristic p. For positive integers \(n_1\), \(n_2\), and \(n_3\) satisfying \(n_1 \le n_2 \le n_3\), the Jordan canonical form of the \(n_1 n_2 n_3 \times n_1 n_2 n_3\) matrix \(J_{n_1} \otimes J_{n_2} \otimes J_{n_3}\) has the form \(J_{\lambda _1} \oplus J_{\lambda _2} \oplus \cdots \oplus J_{\lambda _m}\) where \(\lambda _1 \ge \lambda _2 \ge \cdots \ge \lambda _m>0\) and \(\sum _{i=1}^m \lambda _i=n_1 n_2 n_3\). The partition \(\lambda (n_1,n_2,n_3:p)=(\lambda _1, \lambda _2,\ldots , \lambda _m)\) of \(n_1 n_2 n_3\), which depends on \(n_1\), \(n_2\), \(n_3\), and p, will be called a Jordan partition. We will define what we mean by a standard Jordan partition and give necessary and sufficient conditions for its existence.     

No Outliers in the Spectrum of the Product of Independent Non-Hermitian Random Matrices with Independent Entries

We consider products of independent square random non-Hermitian matrices. More precisely, let \(n\ge 2\) and let \(X_1,\ldots ,X_n\) be independent \(N\times N\) random matrices with independent centered entries (either real or complex with independent real and imaginary parts) with variance \(N^{-1}\). In Götze and Tikhomirov (On the asymptotic spectrum of products of independent random matrices, 2011. arXiv:1012.2710) and O’Rourke and Soshnikov (Electron J Probab 16(81):2219–2245, 2011) it was shown that the limit of the empirical spectral distribution of the product \(X_1\cdots X_n\) is supported in the unit disk. We prove that if the entries of the matrices \(X_1,\ldots ,X_n\) satisfy uniform subexponential decay condition, then the spectral radius of \(X_1\cdots X_n\) converges to 1 almost surely as \(N\rightarrow \infty \).     

Convergence of empirical spectral distributions of large dimensional quaternion sample covariance matrices

In this paper, we establish the limit of empirical spectral distributions of quaternion sample covariance matrices. Motivated by Bai and Silverstein (Spectral analysis of large dimensional random matrices, Springer, New York, 2010) and Mar?enko and Pastur (Matematicheskii Sb, 114:507–536, 1967), we can extend the results of the real or complex sample covariance matrix to the quaternion case. Suppose \(\mathbf X_n = ({x_{jk}^{(n)}})_{p\times n}\) is a quaternion random matrix. For each \(n\), the entries \(\{x_{ij}^{(n)}\}\) are independent random quaternion variables with a common mean \(\mu \) and variance \(\sigma ^2>0\). It is shown that the empirical spectral distribution of the quaternion sample covariance matrix \(\mathbf S_n=n^{-1}\mathbf X_n\mathbf X_n^*\) converges to the Mar?enko–Pastur law as \(p\rightarrow \infty \), \(n\rightarrow \infty \) and \(p/n\rightarrow y\in (0,+\infty )\).     

Newforms with rational coefficients

We consider the set of classical newforms with rational coefficients and no complex multiplication. We study the distribution of quadratic twist-classes of these forms with respect to weight k and minimal level N. We conjecture that for each weight \(k \ge 6\), there are only finitely many classes. In large weights, we make this conjecture effective: in weights \(18 \le k \le 24\), all classes have \(N \le 30\); in weights \(26 \le k \le 50\), all classes have \(N \in \{2,6\}\); and in weights \(k \ge 52\), there are no classes at all. We study some of the newforms appearing on our conjecturally complete list in more detail, especially in the cases \(N=2\), 3, 4, 6, and 8, where formulas can be kept nearly as simple as those for the classical case \(N=1\).     

Computing the Permanent of (Some) Complex Matrices

We present a deterministic algorithm, which, for any given \(0< \epsilon < 1\) and an \(n \times n\) real or complex matrix \(A=\left( a_{ij}\right) \) such that \(\left| a_{ij}-1 \right| \le 0.19\) for all \(i, j\) computes the permanent of \(A\) within relative error \(\epsilon \) in \(n^{O\left( \ln n -\ln \epsilon \right) }\) time. The method can be extended to computing hafnians and multidimensional permanents.     

Quadratic properties of least-squares solutions of linear matrix equations with statistical applications

Assume that a quadratic matrix-valued function \(\psi (X) = Q - X^{\prime }PX\) is given and let \(\mathcal{S} = \left\{ X\in {\mathbb R}^{n \times m} \, | \, \mathrm{trace}[\,(AX - B)^{\prime }(AX - B)\,] = \min \right\} \) be the set of all least-squares solutions of the linear matrix equation \(AX = B\). In this paper, we first establish explicit formulas for calculating the maximum and minimum ranks and inertias of \(\psi (X)\) subject to \(X \in {\mathcal S}\), and then derive from the formulas the analytic solutions of the two optimization problems \(\psi (X) =\max \) and \(\psi (X)= \min \) subject to \(X \in \mathcal{S}\) in the Löwner partial ordering. As applications, we present a variety of results on equalities and inequalities of the ordinary least squares estimators of unknown parameter vectors in general linear models.     

A Quadratically Convergent Algorithm for Structured Low-Rank Approximation

Structured Low-Rank Approximation is a problem arising in a wide range of applications in Numerical Analysis and Engineering Sciences. Given an input matrix \(M\), the goal is to compute a matrix \(M'\) of given rank \(r\) in a linear or affine subspace \(E\) of matrices (usually encoding a specific structure) such that the Frobenius distance \(\left\| M-M' \right\| \) is small. We propose a Newton-like iteration for solving this problem, whose main feature is that it converges locally quadratically to such a matrix under mild transversality assumptions between the manifold of matrices of rank \(r\) and the linear/affine subspace \(E\). We also show that the distance between the limit of the iteration and the optimal solution of the problem is quadratic in the distance between the input matrix and the manifold of rank \(r\) matrices in \(E\). To illustrate the applicability of this algorithm, we propose a Maple implementation and give experimental results for several applicative problems that can be modeled by Structured Low-Rank Approximation: univariate approximate GCDs (Sylvester matrices), low-rank matrix completion (coordinate spaces) and denoising procedures (Hankel matrices).     

On algebras generated by positive operators

We study algebras generated by positive matrices, i.e., matrices with nonnegative entries. Some of our results hold in more general setting of vector lattices. We reprove and extend some theorems that have been recently shown by Kandi? and ?ivic. In particular, we give a more transparent proof of their result that the unital algebra generated by positive idempotent matrices E and F such that \(E F \ge F E\) is equal to the linear span of the set \(\{I, E, F, E F, F E, E F E, F E F, (E F)^2, (F E)^2\}\), and so its dimension is at most 9. We give examples of two positive idempotent matrices that generate unital algebra of dimension 2n if n is even, and of dimension \((2n - 1)\) if n is odd. We also prove that the algebra generated by positive matrices \(B_1\), \(B_2, \ldots , B_k\) is triangularizable if \(A B_i \ge B_i A\) (\(i=1,2, \ldots , k\)) for some positive matrix A with distinct eigenvalues.