Convergence in Hilbert's metric and convergence in direction

Hilbert's metric on a cone K is a measure of distance between the rays of K. Hilbert's metric has many applications, but they all depend on the equivalence between closeness of two rays in the Hilbert metric and closeness of the two unit vectors along these rays (in the usual sense). A necessary and sufficient condition on K for this equivalence to hold is given.     

Incidence matrices of t-designs

We point out a generalization of the matrix equation NN^T=(r? λ)I+λJ to t-designs with t>2 and derive extensions of Fisher's, Connor's, and Mann's inequalities for block designs.     

A characterization of those categories whose internal logic is Hilbert's ε-calculus

We characterize categories whose internal logic is Hilbert's ε-calculus as those categories which have a proper factorization system satisfying the axiom of choice and in which every non-initial object is injective. We provide an example of such a category where the law of excluded middle is not valid.     

Integer programming formulation of combinatorial optimization problems

This paper considers in a somewhat general setting when a combinatorial optimization problem can be formulated as an (all-integer) integer programming (IP) problem. The main result is that any combinatorial optimization problem can be formulated as an IP problem if its feasible region S is finite but there are many rather sample problems that have no IP formulation if their S is infinite. The approach used for finite S usually gives a formulation with a relatively small number of additional variables for example, an integer polynomial of n 0?1 variables requires at most n + 1 additional variables by our approach, whereas 2ⁿ - (n + 1) additional variables at maximum are required by other existing methods. Finally, the decision problem of deciding whether an arbitrarily given combinatorial optimization problem has an IP formulation is considered and it is shown by an argument closely related to Hilbert's tenth problem (drophantine equations) that no such algorithm exists.     

Stochastic processes and special functions: On the probabilistic origin of some positive kernels associated with classical orthogonal polynomials

We describe and investigate a class of Markovian models based on a form of “dynamic occupancy problem” originating in statistical mechanics. The most fundamental of these gives rise to a transition-probability matrix over (N + 1) discrete states, which proves to have the Hahn polynomials as eigenvectors. The structure of this matrix, which is a convolution of two negative hypergeometric distributions, leads to a factorization into finite-difference sumoperators having forms analogous to the Erdelyi-Kober operators for the continuous variable. These make possible the exact solution of the corresponding eigenvalue problem and hence the spectral representation of the transition matrix. By taking suitable limits, further families of Markov processes can be generated having other classical polynomials as eigenvectors; these, like the polynomials, inherit their properties from the original Hahn system. The Meixner, Jacobi and Laguerre systems arise in this way, having their origin in variants of the basic model. In the last of these cases, the spectral resolution of the continuous transition kernel proves to be identical with Erdelyi's (1938) bilinear formula, which is thus both generalized and given a physical interpretation. Various symmetry and “duality” properties are explored and a number of interesting formulas are obtained as by-products. The use of statistical models to generate kernels, which are thereby guaranteed to be both positive and positive definite, appears to be mathematically fruitful, while the models themselves seem likely to have application to a variety of topics in applied probability.     

Memory requirements for balanced computer architectures

In this paper, a processing element (PE) is characterized by its computation bandwidth, I/O bandwidth, and the size of its local memory. In carrying out a computation, a PE is said to be balanced if the computing time equals the I/O time. Consider a balanced PE for some computation. Suppose that the computation band-width of the PE is increased by a factor of α relative to its I/O bandwidth. Then when carrying out the same computation the PE will be imbalanced; i.e., it will have to wait for I/O. A standard method of avoiding this I/O bottleneck is to reduce the overall I/O requirement of the PE by increasing the size of its local memory. This paper addresses the question of by how much the PE's local memory must be enlarged in order to restore balance.The following results are shown: For matrix computations such as matrix multiplication and Gaussian elimination, the size of the local memory must be increased by a factor of α². For computations such as relaxation on a k-dimensional grid, the local memory must be enlarged by a factor of α^k. For some other computations such as the FFT and sorting, the increase is exponential; i.e., the size of the new memory must be the size of the original memory to the αth power. All these results indicate that to design a balanced PE, the size of its local memory must be increased much more rapidly than its computation bandwidth. This phenomenon seems to be common for many computations where an output may depend on a large subset of the inputs.Implications of these results for some parallel computer architectures are also discussed. One particular result is that to balance an array of p linearly connected PEs for performing matrix computations such as matrix multiplication and matrix triangularization, the size of each PE's local memory must grow linearly with p. Thus, the larger the array is, the larger each PE's local memory must be.     

Compositional complexity of Boolean functions

We define two measures, γ and c, of complexity for Boolean functions. These measures are related to issues of functional decomposition which (for continuous functions) were studied by Arnol'd, Kolmogorov, Vitu?kin and others in connection with Hilbert's 13th Problem. This perspective was first applied to Boolean functions in [1]. Our complexity measures differ from those which were considered earlier [3, 5, 6, 9, 10] and which were used by Ehrenfeucht and others to demonstrate the great complexity of most decision procedures. In contrast to other measures, both γ and c (which range between 0 and 1) have a more combinatorial flavor and it is easy to show that both of them are close to 0 for literally all “meaningful” Boolean functions of many variables. It is not trivial to prove that there exist functions for which c is close to 1, and for γ the same question is still open. The same problem for all traditional measures of complexity is easily resolved by statistical considerations.     

A personnel assignment problem

The following personnel assignment problem is considered. Let (T, ?) be a linearly ordered set where T is a set (of people), and let (P, ?) be a partially ordered set where P, a set of positions of two types, is of the same cardinality as T. Each person i in T is to be assigned to a position. A feasible assignment of personnel to positions is an embedding of (P, ?) in (T, ?). Given measures of each person's effectiveness in both types of positions, an optimal assignment maximizes the total measure of effectiveness. The general assignment problem is shown to be NP-complete. O(n log n) algorithms for two special cases of the problem are presented.     

Pairs of alternating forms and products of two skew-symmetric matrices

Let k be a field, and let S,T,S₁,T₁ be skew-symmetric matrices over k with S,S₁ both nonsingular (if k has characteristic 2, a skew-symmetric matrix is a symmetric one with zero diagonal). It is shown that there exists a nonsingular matrix P over k with P'SP = S₁, P'TP = T₁ (where P' denotes the transpose of P) if and only if S^-1T and S^-1₁T₁ are similar. It is also shown that a 2m×2m matrix over k can be factored as ST, with S,T skew-symmetric and S nonsingular, if and only if A is similar to a matrix direct sum B⊕B where B is an m×m matrix over k. This is equivalent to saying that all elementary divisors of A occur with even multiplicity. An extension of this result giving necessary and sufficient conditions for a square matrix to be so expressible, without assuming that either S or T is nonsingular, is included.     

Computation of roots of real and complex matrices

An algorithm is presented in this paper by which the rth root of real or complex matrices can be found without the computation of the eigenvalues and eigenvectors of the matrix. All required computations are in the real domain. The method is based on the Newton-Raphson algorithm and is capable of finding roots even when the matrix is defective. Computing the root of a matrix from eigenvalues and eigenvectors would be the preferred method if these data were available.     

Jordan-Ketten und realisierungen rationaler matrizen

If the inverse of a square polynomial matrix L(s) is proper rational, then L(s)^-1 can be written as C(sI–J)^-1B. The result of this note states that if J is an nXn Jordan matrix, with n=degreedetL(s), then C consists of Jordan chains of L(s), and B^T of Jordan chains of L(s)^T. This is a generalization of the fact that each matrix which transforms a complex matrix A into Jordan form is made up of eigenvectors and generalized eigenvectors of A. The proof of our result relies on the realization theory of rational matrices.     

A note on the eigenvectors of Hadamard matrices of order 2n

A set of 2ⁿ mutually orthogonal eigenvectors of Hadamard matrices of order 2ⁿ are explicitly derived, and their properties are developed. An instance of Pell's equation driven by the Thue-Morse sequence is noted.     

On the eigenvalues of specially low-rank perturbed matrices

We study the eigenvalues of a matrix A perturbed by a few special low-rank matrices. The perturbation is constructed from certain basis vectors of an invariant subspace of A, such as eigenvectors, Jordan vectors, or Schur vectors. We show that most of the eigenvalues of the low-rank perturbed matrix stayed unchanged from the eigenvalues of A; the perturbation can only change the eigenvalues of A that are related to the invariant subspace. Existing results mostly studied using eigenvectors with full column rank for perturbations, we generalize the results to more general settings. Applications of our results to a few interesting problems including the Google’s second eigenvalue problem are presented.     

Score vectors of tournaments

A tournament T on any set X is a dyadic relation such that for any x, y ∈ X (a) (x, x) ? T and (b) if x ≠ y then (x, y) ∈ T iff (y, x) ? T. The score vector of T is the cardinal valued function defined by R(x) = |{y ∈ X : (x, y) ∈ T}|. We present theorems for infinite tournaments analogous to Landau's necessary and sufficient conditions that a vector be the score vector for some finite tournament. Included also is a new proof of Landau's theorem based on a simple application of the “marriage” theorem.     

Approximation to Complex numbers by certain biquadratic numbers

A result of Davenport and Schmidt related to Wirsing's problem is generalized so that complex numbers are approximated using the algebraic integers T from the field, $\text{Q((?D)}{}^{\mathrm{<mtext>1</mtext><mtext>2</mtext>}}\text{)}$, together with numbers of degree two over T. Here D is a square free positive integer.     

Eigenvectors of a kurtosis matrix as interesting directions to reveal cluster structure

In this paper we study the properties of a kurtosis matrix and propose its eigenvectors as interesting directions to reveal the possible cluster structure of a data set. Under a mixture of elliptical distributions with proportional scatter matrix, it is shown that a subset of the eigenvectors of the fourth-order moment matrix corresponds to Fisher’s linear discriminant subspace. The eigenvectors of the estimated kurtosis matrix are consistent estimators of this subspace and its calculation is easy to implement and computationally efficient, which is particularly favourable when the ratio n/p is large.     

Scheduling transmissions in a network

Consider k stations wishing to transmit messages over a network of channels to a common receiver. The capacity of a channel is the maximum amount of data which can be transmitted in a time unit. In addition to the transmission stations, the network contains switching nodes. Given that the jth station has σ_j messages (j = 1,…, k) to transmit, it is desired to find a schedule with minimum completion time T. The amount of data sent over a channel may vary in time. A schedule is stationary if the amount of data sent in a time unit is constant throughout the schedule. It is first shown that for every schedule there exists a stationary schedule with the same completion time. Thus, the search for an optimum schedule is confined to stationary schedules. The problem of finding an optimum stationary schedule is formulated as a multisource single-sink network flow problem, in which the net amount of outgoing flow from each source (transmission station) within one time unit is $\text{σ}{}_{\mathrm{j}}\text{T}$. An O(k|E||V|²) time algorithm to find the minimum T similar to Dinic's flow algorithm is suggested. Using Sleator and Tarjan's techniques an O(k²|E||V|log|V|) algorithm is derived. The running time of both algorithms is independent of the σ_j's and the capacities.     

Some angularity and inertia theorems related to normal matrices

Motivated by the definition of the inertia, introduced by Ostrowski and Schneider, a notion of angularity of a matrix is defined. The angularity characterizes the distribution of arguments of eigenvalues of a matrix. It is proved that if B and C are nonsingular matrices, then B¹AB and C¹AC have the same angularity provided they are diagonal. Some well-known inertia theorems (e.g. Sylvester's law) have been deduced as corollaries of this result. The case when C is permitted to be singular is discussed next. Finally, we prove that (a) any linear transformation T, on the set of n by n complex matrices, mapping Hermitian matrices into themselves and preserving the inertia of each Hermitian matrix is of the form T(A)=C¹AC or T(A)=C^1LA′C where C is some nonsingular matrix, and (b) any linear transformation T mapping normal matrices into normal matrices and preserving the angularity of each normal matrix is also of one of the above forms, but with C=kU where k≠0 and U is unitary.     

Matroids,generalized networks,and electric network synthesis

Matroid theory has been applied to solve problems in generalized assignment, operations research, control theory, network theory, flow theory, generalized flow theory or linear programming, coding theory, and telecommunication network design. The operations of matroid union, matroid partitioning, matroid intersection, and the theorem on the greedy algorithm, Rado's theorem, and Brualdi's symmetric version of Rado's theorem have been important for some of these applications. In this paper we consider the application of matroids to solve problems in network synthesis. Previously Bruno and Weinberg defined a generalized network, which is a network based on a matroid rather than a graph; for a generalized network the duality principle holds whereas it does not hold for a network based on a graph. We use the concept of the generalized network to formulate a solution to the following problem: What are the necessary and sufficient conditions for a singular matrix of real numbers, of order p and rank s, to be realizable as the open-circuit resistance matrix of a resistance p-port network. A simple algorithm is given for carriyng out the synthesis. We then present a number of unsolved problems, included among which is what could be called the four-color problem of network synthesis, namely, the resistance n-port problem.     

The diagonal hypergraph of a matrix (bipartite graph)

Let A be an n x n matrix of 0's and 1's (a bipartite graph). The diagonal hypergraph of A is the hypergraph whose vertices correspond to the 1's (edges) of A and whose edges correspond to the positive diagonals (1-factors) of A. The numerical invariants of this hypergraph are investigated.