Performance of the Gibbs,Hit-and-Run,and Metropolis Samplers

Abstract

We consider the performance of three Monte Carlo Markov-chain samplers—the Gibbs sampler, which cycles through coordinate directions; the Hit-and-Run (H&R) sampler, which randomly moves in any direction; and the Metropolis sampler, which moves with a probability that is a ratio of likelihoods. We obtain several analytical results. We provide a sufficient condition of the geometric convergence on a bounded region S for the H&R sampler. For a general region S, we review the Schervish and Carlin sufficient geometric convergence condition for the Gibbs sampler. We show that for a multivariate normal distribution this Gibbs sufficient condition holds and for a bivariate normal distribution the Gibbs marginal sample paths are each an AR(1) process, and we obtain the standard errors of sample means and sample variances, which we later use to verify empirical Monte Carlo results. We empirically compare the Gibbs and H&R samplers on bivariate normal examples. For zero correlation, the Gibbs sampler provides independent data, resulting in better performance than H&R. As the absolute value of the correlation increases, H&R performance improves, with H&R substantially better for correlations above .9. We also suggest and study methods for choosing the number of replications, for estimating the standard error of point estimators and for reducing point-estimator variance. We suggest using a single long run instead of using multiple iid separate runs. We suggest using overlapping batch statistics (obs) to get the standard errors of estimates; additional empirical results show that obs is accurate. Finally, we review the geometric convergence of the Metropolis algorithm and develop a Metropolisized H&R sampler. This sampler works well for high-dimensional and complicated integrands or Bayesian posterior densities.     

Slanted matrices, Banach frames, and sampling

In this paper we present a rare combination of abstract results on the spectral properties of slanted matrices and some of their very specific applications to frame theory and sampling problems. We show that for a large class of slanted matrices boundedness below of the corresponding operator in ?p for some p implies boundedness below in ?p for all p. We use the established result to enrich our understanding of Banach frames and obtain new results for irregular sampling problems. We also present a version of a non-commutative Wiener's lemma for slanted matrices.     

Lorentz-Covariant Ultradistributions, Hyperfunctions, and Analytic Functionals

We generalize the theory of Lorentz-covariant distributions to broader classes of functionals including ultradistributions, hyperfunctions, and analytic functionals with a tempered growth. We prove that Lorentz-covariant functionals with essential singularities can be decomposed into polynomial covariants and establish the possibility of the invariant decomposition of their carrier cones. We describe the properties of odd highly singular generalized functions. These results are used to investigate the vacuum expectation values of nonlocal quantum fields with an arbitrary high-energy behavior and to extend the spin–statistics theorem to nonlocal field theory.     

Subcompact cardinals, squares, and stationary reflection

We generalise Jensen’s result on the incompatibility of subcompactness with □. We show that α ⁺-subcompactness of some cardinal less than or equal to α precludes ${\square _\alpha }$ , but also that square may be forced to hold everywhere where this obstruction is not present. The forcing also preserves other strong large cardinals. Similar results are also given for stationary reflection, with a corresponding strengthening of the large cardinal assumption involved. Finally, we refine the analysis by considering Schimmerling’s hierarchy of weak squares, showing which cases are precluded by α ⁺-subcompactness, and again we demonstrate the optimality of our results by forcing the strongest possible squares under these restrictions to hold.     

Backward errors for eigenvalues and eigenvectors of Hermitian,skew-Hermitian,H-even and H-odd matrix polynomials

We discuss the perturbation analysis for eigenvalues and eigenvectors of structured homogeneous matrix polynomials with Hermitian, skew-Hermitian, H-even and H-odd structure. We construct minimal structured perturbations (structured backward errors) such that an approximate eigenvalue and eigenvector pair (finite or infinite eigenvalues) is an exact eigenvalue eigenvector pair of an appropriately perturbed structured matrix polynomial. We present various comparisons with unstructured backward errors and previous backward errors constructed for the non-homogeneous case and show that our results generalize previous results.     

Disjoint paths, planarizing cycles, and spanning walks

We study the existence of certain disjoint paths in planar graphs and generalize a theorem of Thomassen on planarizing cycles in surfaces. Results are used to prove that every 5-connected triangulation of a surface with sufficiently large representativity is hamiltonian, thus verifying a conjecture of Thomassen. We also obtain results about spanning walks in graphs embedded in a surface with large representativity.

     

Divisors on graphs,binomial and monomial ideals,and cellular resolutions

We study various binomial and monomial ideals arising in the theory of divisors, orientations, and matroids on graphs. We use ideas from potential theory on graphs and from the theory of Delaunay decompositions for lattices to describe their minimal polyhedral cellular free resolutions. We show that the resolutions of all these ideals are closely related and that their \({\mathbb {Z}}\)-graded Betti tables coincide. As corollaries, we give conceptual proofs of conjectures and questions posed by Postnikov and Shapiro, by Manjunath and Sturmfels, and by Perkinson, Perlman, and Wilmes. Various other results related to the theory of chip-firing games on graphs also follow from our general techniques and results.     

Subnormal operators, self-commutators, and pseudocontinuations

We study pure subnormal operators whose self-commutators have zero as an eigenvalue. We show that various questions in this are closely related to questions involving approximation by functions satisfying and to the study ofgeneralized quadrature domains.First some general results are given that apply to all subnormal operators within this class; then we consider characterizing the analytic Toeplitz operators, the Hardy operators and cyclic subnormal operators whose self-commutators have zero as an eigenvalue.     

A bijection for triangulations, quadrangulations, pentagulations, etc.

A d-angulation is a planar map with faces of degree d. We present for each integer d?3 a bijection between the class of d-angulations of girth d (i.e., with no cycle of length less than d) and a class of decorated plane trees. Each of the bijections is obtained by specializing a “master bijection” which extends an earlier construction of the first author. Our construction unifies known bijections by Fusy, Poulalhon and Schaeffer for triangulations (d=3) and by Schaeffer for quadrangulations (d=4). For d?5, both the bijections and the enumerative results are new.We also extend our bijections so as to enumerate p-gonal d-angulations (d-angulations with a simple boundary of length p) of girth d. We thereby recover bijectively the results of Brown for simple p-gonal triangulations and simple 2p-gonal quadrangulations and establish new results for d?5.A key ingredient in our proofs is a class of orientations characterizing d-angulations of girth d. Earlier results by Schnyder and by De Fraysseix and Ossona de Mendez showed that simple triangulations and simple quadrangulations are characterized by the existence of orientations having respectively indegree 3 and 2 at each inner vertex. We extend this characterization by showing that a d-angulation has girth d if and only if the graph obtained by duplicating each edge d−2 times admits an orientation having indegree d at each inner vertex.     

Unitary local systems, multiplier ideals, and polynomial periodicity of Hodge numbers

The space of unitary local systems of rank one on the complement of an arbitrary divisor in a complex projective algebraic variety can be described in terms of parabolic line bundles. We show that multiplier ideals provide natural stratifications of this space. We prove a structure theorem for these stratifications in terms of complex tori and convex rational polytopes, generalizing to the quasi-projective case results of Green-Lazarsfeld and Simpson. As an application we show the polynomial periodicity of Hodge numbers hq,0 of congruence covers in any dimension, generalizing results of E. Hironaka and Sakuma. We extend the structure theorem and polynomial periodicity to the setting of cohomology of unitary local systems. In particular, we obtain a generalization of the polynomial periodicity of Betti numbers of unbranched congruence covers due to Sarnak-Adams. We derive a geometric characterization of finite abelian covers, which recovers the classic one and the one of Pardini. We use this, for example, to prove a conjecture of Libgober about Hodge numbers of abelian covers.     

Higher-dimensional orders,graph-orders,and derived equivalences

Green-orders (tree-orders) in the classical one-dimensional case are the setting, to understand p-adic blocks with cyclic defect of finite groups. Blocks with “cyclic defect” of Hecke orders however, are Green-orders over two-dimensional rings. Hecke orders of dihedral groups of order divisible by 4 are even defined over a three-dimensional ring. We extend the notion of Green-orders to orders associated to a locally embedded graph instead of a tree, and to general complete regular local noetherian ground rings of finite dimension. We extend the result, that classical tree-orders are derived equivalent to star-orders. We then use these results to clarify the derived equivalence classes of tame algebras of Dihedral type.     

Ramsey property, ultrametric spaces, finite posets, and universal minimal flows

We introduce the class COU _S of finite ultrametric spaces with distances in the set S and with two additional linear orderings. We also introduce the class EOP of finite posets with two additional linear orderings. In this paper, we prove that COU _S and EOP are Ramsey classes. In addition, we give an application of our results to calculus of universal minimal flows.     

Minimal enclosing discs,circumcircles, and circumcenters in normed planes (Part II)

Until now there are almost no results on the precise geometric location of minimal enclosing balls of simplices in finite-dimensional real Banach spaces. We give a complete solution of the two-dimensional version of this problem, namely to locate minimal enclosing discs of triangles in arbitrary normed planes. It turns out that this solution is based on the classification of all possible shapes that the intersection of two norm circles can have, and on a new classification of triangles in normed planes via their angles. We also mention that our results are closely related to basic notions like coresets, Jung constants, the monotonicity lemma, and d-segments.     

On convex modified output least-squares for elliptic inverse problems: stability,regularization, applications,and numerics

Abstract

Inverse problems of identifying parameters in partial differential equations constitute an important class of problems with diverse real-world applications. These identification problems are commonly explored in an optimization framework and there are many optimization formulations having their own advantages and disadvantages. Although a non-convex output least-squares (OLS) objective is commonly used, a convex-modified output least-squares (MOLS) has shown encouraging results in recent years. In this work, we focus on various aspects of the MOLS approach. We devise a rigorous (quadratic and non-quadratic) regularization framework for the identification of smooth as well as discontinuous coefficients. This framework subsumes the total variation regularization that has attracted a great deal of attention in identifying sharply varying coefficients and also in image processing. We give new existence results for the regularized optimization problems for OLS and MOLS. Restricting to the Tikhonov (quadratic) regularization, we carry out a detailed study of various stability aspects of the inverse problem under data perturbation and give new stability estimates for general inverse problems using OLS and MOLS formulations. We give a discretization scheme for the continuous inverse problem and prove the convergence of the discrete inverse problem to the continuous one. We collect discrete formulas for OLS and MOLS and compute their gradients and Hessians. We present applications of our theoretical results. To show the feasibility of the MOLS framework, we also provide computational results for the inverse problem of identifying parameters in three different classes of partial differential equations .     

New distribution formulas for classical gas, clusters, and phase transitions

We obtain a Bose-Einstein-type distribution for the classical vapor. We show that the analogue of the Bose condensate is the formation of clusters. We write a new PV diagram for interaction with the form of the Lennard-Jones potential using scattering theory and a strict constraint. We compare our results with experimental data. Dedicated to the memory of a dear friend, Tolya Karatsuba __________ Translated from Teoreticheskaya i Matematicheskaya Fizika, Vol. 157, No. 2, pp. 250–272, November, 2008.     

Cayley compositions,partitions, polytopes,and geometric bijections

In 1857, Cayley showed that certain sequences, now called Cayley compositions, are equinumerous with certain partitions into powers of 2. In this paper we give a simple bijective proof of this result and a geometric generalization to equality of Ehrhart polynomials between two convex polytopes. We then apply our results to give a new proof of Braun?s conjecture proved recently by the authors [15].     

A compactness theorem for singular cardinals,free algebras,Whitehead problem and tranversals

We prove, in an axiomatic way, a compactness theorem for singular cardinals. We apply it to prove that, for singular λ, every λ-free algebra is free; and similar compactness results for transversals and colouring numbers. For the general result on free algebras, we develop some filters onS _k(A). As an application we conclude thatV=L implies that every Whitehead group is free.     

Sagitta,Lenses, and Maximal Volume

We give a characterization of critical points that allows us to define a metric invariant on all Riemannian manifolds M with a lower sectional curvature bound and an upper radius bound. We show there is a uniform upper volume bound for all such manifolds with an upper bound on this invariant. We generalize results by Grove and Petersen by showing any such M that has volume sufficiently close to this upper bound is homeomorphic to the standard sphere \(S^{n}\) or a standard lens space \(S^n/{\mathbb {Z}}_m\) where \(m\in \{2,3,\ldots \}\) is no larger than an a priori constant.     

Base sequences,complementary ternary sequences,and orthogonal designs

We construct several new families of orthogonal designs related to one of the weighing matrix conjectures. We also construct some new complementary quadruples (A; B; C; D) of {0, ± 1}-sequences of length n and total weight w, denoted as CTQ(n, w). The word complementary refers to the fact that the sum of nonperiodic autocorrelation functions of these four sequences is a delta function. We propose a conjectural formula for n(w), the smallest positive integer n such that CTQ(n, w) exist, and survey the known results for all w ≤ 100. We also exhibit the first example of base sequences BS(32, 31). © 1996 John Wiley & Sons, Inc.