DOCSLIB.ORG

The Use of the Vec-Permutation Matrix in Spatial Matrix Population Models Christine M

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
The Use of the Vec-Permutation Matrix in Spatial Matrix Population Models Christine M Ecological Modelling 188 (2005) 15–21 The use of the vec-permutation matrix in spatial matrix population models Christine M. Hunter∗, Hal Caswell Biology Department MS-34, Woods Hole Oceanographic Institution, Woods Hole, MA 02543, USA Abstract Matrix models for a metapopulation can be formulated in two ways: in terms of the stage distribution within each spatial patch, or in terms of the spatial distribution within each stage. In either case, the entries of the projection matrix combine demographic and dispersal information in potentially complicated ways. We show how to construct such models from a simple block-diagonal formulation of the demographic and dispersal processes, using a special permutation matrix called the vec-permutation matrix. This formulation makes it easy to calculate and interpret the sensitivity and elasticity of λ to changes in stage- and patch-specific demographic and dispersal parameters. © 2005 Christine M. Hunter and Hal Caswell. Published by Elsevier B.V. All rights reserved. Keywords: Matrix population model; Spatial model; Metapopulation; Multiregional model; Dispersal; Movement; Sensitivity; Elasticity 1. Introduction If each local population is described by a stage- classified matrix population model (see Caswell, A spatial population model describes a finite set of 2001), a spatial matrix model describing the metapop- discrete local populations coupled by the movement ulation can be written as or dispersal of individuals. Human demographers call these multiregional populations (Rogers, 1968, 1995), n(t + 1) = An(t) (1) ecologists call them multisite populations (Lebreton, 1996) or metapopulations (Hanski, 1999). The dynam- The population vector, n, includes the densities of ics of metapopulations are determined by the patterns each stage in each local population (which we will of dispersal among the local populations and the de- refer to here as a patch). The projection matrix, A, mographic conditions experienced by the local popu- includes both demographic processes, (which gener- lations. ally differ among patches) and dispersal processes (which generally differ among stages). The asymp- λ ∗ Corresponding author. Tel.: +1 508 289 3245; totic population growth rate, , is the dominant fax: +1 508 457 2134. eigenvalue of A, and the stable stage × patch dis- E-mail address: [email protected] (C.M. Hunter). tribution and the reproductive values are given by 0304-3800/$ – see front matter © 2005 Christine M. Hunter and Hal Caswell. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.ecolmodel.2005.05.002 16 C.M. Hunter, H. Caswell / Ecological Modelling 188 (2005) 15–21 the corresponding right and left eigenvectors w and ond arrangement (nstages) the subvectors give the spa- v. tial distribution of each stage. Note that in general it is The elements of A are combinations of demographic not possible to write a model in which the matrix N(t) rates and dispersal probabilities. It can be challenging replaces the vector n(t)inEq.(1) and is projected by to formulate these elements, and difficult to analyze multiplication by a matrix (Logofet, 2002). the role of particular parameters. This can be espe- To model demography without dispersal, it would cially problematic for perturbation analyses. The sen- be convenient to organize the population by patches. sitivity and elasticity of λ to the elements of A are Let Bi be the s × s demographic projection matrix for of little interest, because a change in aij has no sim- patch i, such that n·i(t + 1) = Bin·i(t). Then, without ple biological interpretation. Instead, attention focuses dispersal, the metapopulation would be projected by a on the effects of changes in demography and dispersal block diagonal matrix B separately. Here we describe a method for constructing spa- B1 0 ··· 0 tial matrix population models that keeps demographic n· n· 1 ··· 1 and dispersal parameters clearly distinguished and sep- 0 B2 0 . t + = . t arates their effects in perturbation analysis. It permits . ( 1) . . ( ) .. the inclusion of complicated seasonal patterns of dis- n·p n·p 00··· Bp persal. We describe model construction and the formu- lation of sensitivity and elasticity analyses, and apply B the method to two simple examples. (4) To model dispersal without demography, it would be 2. Spatial model construction more convenient to organize the population by stages. Let Mh be the p × p dispersal matrix for stage h,so Constructing a spatial matrix model requires speci- (h) that m is the probability that an individual of stage fying the state of the metapopulation, the demographic ij h moves from patch j to patch i. Then dispersal would characteristics of each patch, and the dispersal of indi- be described by a block diagonal matrix M viduals among the patches. To describe the state of the metapopulation let p be the number of patches and s be ··· T M1 0 0 T the number of stages. The state of the metapopulation n1· n1· 0 M2 ··· 0 can then be described by the matrix . t + = . t . ( 1) . . ( ) .. n11 n12 ··· n1p T T ns· ns· n n ··· n 00··· Ms 21 22 2p t = t N( ) . ( ) (2) M . .. (5) ns1 ns2 ··· nsp Constructing a spatial model that incorporates both de- where nij (t) is the density of stage i in patch j at time t. The population vector n in (1) can be written in two mography and dispersal has previously required sacri- ficing one or both of these block-diagonal forms. Our ways. Let ni· =rowi and n·j = column j of N, then goal is a method of model construction that maintains T them both. n·1 n1· As many other authors have done, we assume that = . = . npatches . or nstages . (3) demography and dispersal can be treated as operat- ing sequentially within the projection interval (e.g., n·p nT s· Hastings, 1992; Bravo de la Parra et al., 1995, 1997; In the first arrangement (npatches), the subvectors give Sanz and Bravo de la Parra, 1999, 2000; Lebreton and the stage distribution within each patch. In the sec- Gonzalez-Davila, 1993), although either process may C.M. Hunter, H. Caswell / Ecological Modelling 188 (2005) 15–21 17 be seasonal and they may occur in any order. Then, Table 1 constructing a model that maintains the block diagonal The projection matrix A and the sensitivity matrices SB and SM as forms of the demography and dispersal matrices in (4) a function of the arrangement of the population vector (by patches or by stages) and the order of demography and dispersal within the and (5) requires converting a population vector orga- projection interval nized by patches into a population vector organized by Population vector arrangement stages and vice versa. This conversion is accomplished By patches By stages by noting that Demography then dispersal T T = (a) A = P MPB (b) A = M P B P npatches vec(N) (6) T T T T SB = P M PSA SB = P M SA P = BT T = BT T T SM PSA P SM SAP P nstages = vec(N ) (7) Dispersal then demography where the vec operator, vec(·), stacks the columns of (c) A = B PT M P (d) A = P B PT M = T MT = T MT a matrix one on top of the other (this is implemented SB SA P PSB P SA P SM = P BT S PT SM = P BT PT S with the command N(:) in MATLAB). The vectors A A (6) and (7) are related by a special permutation matrix, In all cases, the elasticity matrices satisfy EB = B ◦ SB/λ and EM = M ◦ /λ P, called the vec-permutation matrix; i.e., SM . vec(NT) = P vec(N) (8) If dispersal is followed by demography and we ar- range n by stages, we would have Henderson and Searle (1981) reviewed the properties and derivation of the vec-permutation matrix. It has T T n1· n1· dimension (sp × sp) and is given by . T . . (t + 1) = P B P M . (t) (11) s p T T T P(s, p) = Eij ⊗ Eij (9) ns· ns· i=1 j=1 In this case, the projection matrix is A = P B PT M. s × p i, j where Eij is an matrix witha1inthe( ) The four possible combinations of arrangements of ⊗ position and zeros elsewhere and denotes the Kro- n and sequences of a single demography and a sin- necker matrix product. As with any permutation matrix, gle dispersal event lead to the four projection matrices T = −1 P P . shown in Table 1. Each of these projection matrices can Suppose that, within each projection interval, de- be obtained from any of the others by a cyclic permu- mographic change occurs within each patch and then tation of the matrices making up the product. Hence dispersal redistributes individuals among patches. If we their eigenvalue spectra are identical (Horn and John- organize n by patches we can use the vec-permutation son, 1990, Theorem 1.3.20; see Caswell, 2001 p. 350), matrix to write λ as are the sensitivities and elasticities of to the entries B M n·1 n·1 of and obtained from each. . t + = T M B . t . ( 1) P P . ( ) (10) n·p n·p 3. Sensitivity analysis The projection matrix A in (1) is then A = PT M P B. Eqs. (10) and (11) describe population growth in a In this case, A first applies the block-demography periodic environment that alternates between episodes matrix B to the population vector organized by of demography and dispersal. The sensitivity and elas- patches, then permutes the result to reorganize the ticity of λ to changes in demography or dispersal at any population vector by stages, applies the block- point in the projection cycle can be calculated using re- dispersal matrix M, and then permutes the result sults of Caswell and Trevisan (1994) and Lesnoff et al.
Load more

Recommended publications
  • Solving the Multi-Way Matching Problem by Permutation Synchronization
    Solving the multi-way matching problem by permutation synchronization Deepti Pachauriy Risi Kondorx Vikas Singhy yUniversity of Wisconsin Madison xThe University of Chicago [email protected] [email protected] [email protected] http://pages.cs.wisc.edu/∼pachauri/perm-sync Abstract The problem of matching not just two, but m different sets of objects to each other arises in many contexts, including finding the correspondence between feature points across multiple images in computer vision. At present it is usually solved by matching the sets pairwise, in series. In contrast, we propose a new method, Permutation Synchronization, which finds all the matchings jointly, in one shot, via a relaxation to eigenvector decomposition. The resulting algorithm is both computationally efficient, and, as we demonstrate with theoretical arguments as well as experimental results, much more stable to noise than previous methods. 1 Introduction 0 Finding the correct bijection between two sets of objects X = fx1; x2; : : : ; xng and X = 0 0 0 fx1; x2; : : : ; xng is a fundametal problem in computer science, arising in a wide range of con- texts [1]. In this paper, we consider its generalization to matching not just two, but m different sets X1;X2;:::;Xm. Our primary motivation and running example is the classic problem of matching landmarks (feature points) across many images of the same object in computer vision, which is a key ingredient of image registration [2], recognition [3, 4], stereo [5], shape matching [6, 7], and structure from motion (SFM) [8, 9]. However, our approach is fully general and equally applicable to problems such as matching multiple graphs [10, 11].
    [Show full text]
  • Arxiv:1912.06217V3 [Math.NA] 27 Feb 2021
    ROUNDING ERROR ANALYSIS OF MIXED PRECISION BLOCK HOUSEHOLDER QR ALGORITHMS∗ L. MINAH YANG† , ALYSON FOX‡, AND GEOFFREY SANDERS‡ Abstract. Although mixed precision arithmetic has recently garnered interest for training dense neural networks, many other applications could benefit from the speed-ups and lower storage cost if applied appropriately. The growing interest in employing mixed precision computations motivates the need for rounding error analysis that properly handles behavior from mixed precision arithmetic. We develop mixed precision variants of existing Householder QR algorithms and show error analyses supported by numerical experiments. 1. Introduction. The accuracy of a numerical algorithm depends on several factors, including numerical stability and well-conditionedness of the problem, both of which may be sensitive to rounding errors, the difference between exact and finite precision arithmetic. Low precision floats use fewer bits than high precision floats to represent the real numbers and naturally incur larger rounding errors. Therefore, error attributed to round-off may have a larger influence over the total error and some standard algorithms may yield insufficient accuracy when using low precision storage and arithmetic. However, many applications exist that would benefit from the use of low precision arithmetic and storage that are less sensitive to floating-point round-off error, such as training dense neural networks [20] or clustering or ranking graph algorithms [25]. As a step towards that goal, we investigate the use of mixed precision arithmetic for the QR factorization, a widely used linear algebra routine. Many computing applications today require solutions quickly and often under low size, weight, and power constraints, such as in sensor formation, where low precision computation offers the ability to solve many problems with improvement in all four parameters.
    [Show full text]
  • Arxiv:1709.00765V4 [Cond-Mat.Stat-Mech] 14 Oct 2019
    Number of hidden states needed to physically implement a given conditional distribution Jeremy A. Owen,1 Artemy Kolchinsky,2 and David H. Wolpert2, 3, 4, ∗ 1Physics of Living Systems Group, Department of Physics, Massachusetts Institute of Technology, 400 Tech Square, Cambridge, MA 02139. 2Santa Fe Institute 3Arizona State University 4http://davidwolpert.weebly.com† (Dated: October 15, 2019) We consider the problem of how to construct a physical process over a finite state space X that applies some desired conditional distribution P to initial states to produce final states. This problem arises often in the thermodynamics of computation and nonequilibrium statistical physics more generally (e.g., when designing processes to implement some desired computation, feedback controller, or Maxwell demon). It was previously known that some conditional distributions cannot be implemented using any master equation that involves just the states in X. However, here we show that any conditional distribution P can in fact be implemented—if additional “hidden” states not in X are available. Moreover, we show that is always possible to implement P in a thermodynamically reversible manner. We then investigate a novel cost of the physical resources needed to implement a given distribution P : the minimal number of hidden states needed to do so. We calculate this cost exactly for the special case where P represents a single-valued function, and provide an upper bound for the general case, in terms of the nonnegative rank of P . These results show that having access to one extra binary degree of freedom, thus doubling the total number of states, is sufficient to implement any P with a master equation in a thermodynamically reversible way, if there are no constraints on the allowed form of the master equation.
    [Show full text]
  • On the Asymptotic Existence of Partial Complex Hadamard Matrices And
    Discrete Applied Mathematics 102 (2000) 37–45 View metadata, citation and similar papers at core.ac.uk brought to you by CORE On the asymptotic existence of partial complex Hadamard provided by Elsevier - Publisher Connector matrices and related combinatorial objects Warwick de Launey Center for Communications Research, 4320 Westerra Court, San Diego, California, CA 92121, USA Received 1 September 1998; accepted 30 April 1999 Abstract A proof of an asymptotic form of the original Goldbach conjecture for odd integers was published in 1937. In 1990, a theorem reÿning that result was published. In this paper, we describe some implications of that theorem in combinatorial design theory. In particular, we show that the existence of Paley’s conference matrices implies that for any suciently large integer k there is (at least) about one third of a complex Hadamard matrix of order 2k. This implies that, for any ¿0, the well known bounds for (a) the number of codewords in moderately high distance binary block codes, (b) the number of constraints of two-level orthogonal arrays of strengths 2 and 3 and (c) the number of mutually orthogonal F-squares with certain parameters are asymptotically correct to within a factor of 1 (1 − ) for cases (a) and (b) and to within a 3 factor of 1 (1 − ) for case (c). The methods yield constructive algorithms which are polynomial 9 in the size of the object constructed. An ecient heuristic is described for constructing (for any speciÿed order) about one half of a Hadamard matrix. ? 2000 Elsevier Science B.V. All rights reserved.
    [Show full text]
  • Triangular Factorization
    Chapter 1 Triangular Factorization This chapter deals with the factorization of arbitrary matrices into products of triangular matrices. Since the solution of a linear n n system can be easily obtained once the matrix is factored into the product× of triangular matrices, we will concentrate on the factorization of square matrices. Specifically, we will show that an arbitrary n n matrix A has the factorization P A = LU where P is an n n permutation matrix,× L is an n n unit lower triangular matrix, and U is an n ×n upper triangular matrix. In connection× with this factorization we will discuss pivoting,× i.e., row interchange, strategies. We will also explore circumstances for which A may be factored in the forms A = LU or A = LLT . Our results for a square system will be given for a matrix with real elements but can easily be generalized for complex matrices. The corresponding results for a general m n matrix will be accumulated in Section 1.4. In the general case an arbitrary m× n matrix A has the factorization P A = LU where P is an m m permutation× matrix, L is an m m unit lower triangular matrix, and U is an×m n matrix having row echelon structure.× × 1.1 Permutation matrices and Gauss transformations We begin by defining permutation matrices and examining the effect of premulti- plying or postmultiplying a given matrix by such matrices. We then define Gauss transformations and show how they can be used to introduce zeros into a vector. Definition 1.1 An m m permutation matrix is a matrix whose columns con- sist of a rearrangement of× the m unit vectors e(j), j = 1,...,m, in RI m, i.e., a rearrangement of the columns (or rows) of the m m identity matrix.
    [Show full text]
  • Think Global, Act Local When Estimating a Sparse Precision
    Think Global, Act Local When Estimating a Sparse Precision Matrix by Peter Alexander Lee A.B., Harvard University (2007) Submitted to the Sloan School of Management in partial fulfillment of the requirements for the degree of Master of Science in Operations Research at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY June 2016 ○c Peter Alexander Lee, MMXVI. All rights reserved. The author hereby grants to MIT permission to reproduce and to distribute publicly paper and electronic copies of this thesis document in whole or in part in any medium now known or hereafter created. Author................................................................ Sloan School of Management May 12, 2016 Certified by. Cynthia Rudin Associate Professor Thesis Supervisor Accepted by . Patrick Jaillet Dugald C. Jackson Professor, Department of Electrical Engineering and Computer Science and Co-Director, Operations Research Center 2 Think Global, Act Local When Estimating a Sparse Precision Matrix by Peter Alexander Lee Submitted to the Sloan School of Management on May 12, 2016, in partial fulfillment of the requirements for the degree of Master of Science in Operations Research Abstract Substantial progress has been made in the estimation of sparse high dimensional pre- cision matrices from scant datasets. This is important because precision matrices underpin common tasks such as regression, discriminant analysis, and portfolio opti- mization. However, few good algorithms for this task exist outside the space of L1 penalized optimization approaches like GLASSO. This thesis introduces LGM, a new algorithm for the estimation of sparse high dimensional precision matrices. Using the framework of probabilistic graphical models, the algorithm performs robust covari- ance estimation to generate potentials for small cliques and fuses the local structures to form a sparse yet globally robust model of the entire distribution.
    [Show full text]
  • Mata Glossary of Common Terms
    Title stata.com Glossary Description Commonly used terms are defined here. Mata glossary arguments The values a function receives are called the function’s arguments. For instance, in lud(A, L, U), A, L, and U are the arguments. array An array is any indexed object that holds other objects as elements. Vectors are examples of 1- dimensional arrays. Vector v is an array, and v[1] is its first element. Matrices are 2-dimensional arrays. Matrix X is an array, and X[1; 1] is its first element. In theory, one can have 3-dimensional, 4-dimensional, and higher arrays, although Mata does not directly provide them. See[ M-2] Subscripts for more information on arrays in Mata. Arrays are usually indexed by sequential integers, but in associative arrays, the indices are strings that have no natural ordering. Associative arrays can be 1-dimensional, 2-dimensional, or higher. If A were an associative array, then A[\first"] might be one of its elements. See[ M-5] asarray( ) for associative arrays in Mata. binary operator A binary operator is an operator applied to two arguments. In 2-3, the minus sign is a binary operator, as opposed to the minus sign in -9, which is a unary operator. broad type Two matrices are said to be of the same broad type if the elements in each are numeric, are string, or are pointers. Mata provides two numeric types, real and complex. The term broad type is used to mask the distinction within numeric and is often used when discussing operators or functions.
    [Show full text]
  • Sparse Precision Matrix Estimation Via Lasso Penalized D-Trace Loss
    Biometrika (2014), 101,1,pp. 103–120 doi: 10.1093/biomet/ast059 Printed in Great Britain Advance Access publication 12 February 2014 Sparse precision matrix estimation via lasso penalized D-trace loss BY TENG ZHANG Department of Mathematics, Princeton University, Fine Hall, Washington Rd, Princeton, Downloaded from New Jersey 08544, U.S.A. [email protected] AND HUI ZOU http://biomet.oxfordjournals.org/ School of Statistics, University of Minnesota, 224 Church St SE, Minneapolis, Minnesota 55455, U.S.A. [email protected] SUMMARY We introduce a constrained empirical loss minimization framework for estimating high- dimensional sparse precision matrices and propose a new loss function, called the D-trace loss, for that purpose. A novel sparse precision matrix estimator is defined as the minimizer of the lasso at University of Minnesota,Walter Library Serial Processing on March 5, 2014 penalized D-trace loss under a positive-definiteness constraint. Under a new irrepresentability condition, the lasso penalized D-trace estimator is shown to have the sparse recovery property. Examples demonstrate that the new condition can hold in situations where the irrepresentability condition for the lasso penalized Gaussian likelihood estimator fails. We establish rates of con- vergence for the new estimator in the elementwise maximum, Frobenius and operator norms. We develop a very efficient algorithm based on alternating direction methods for computing the pro- posed estimator. Simulated and real data are used to demonstrate the computational efficiency of our algorithm and the finite-sample performance of the new estimator. The lasso penalized D-trace estimator is found to compare favourably with the lasso penalized Gaussian likelihood estimator.
    [Show full text]
  • An Efficient ADMM Algorithm for High Dimensional Precision Matrix
    An efficient ADMM algorithm for high dimensional precision matrix estimation via penalized quadratic loss Cheng Wang ∗a and Binyan Jiang yb aSchool of Mathematical Sciences, MOE-LSC,Shanghai Jiao Tong University, Shanghai 200240, China. bDepartment of Applied Mathematics, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong. July 10, 2019 Abstract The estimation of high dimensional precision matrices has been a central topic in statistical learning. However, as the number of parameters scales quadratically with the dimension p, many state-of-the-art methods do not scale well to solve problems with a very large p. In this paper, we propose a very efficient algorithm for precision matrix estimation via penalized quadratic loss functions. Under the high dimension low sample size setting, the computation complexity of our al- gorithm is linear in both the sample size and the number of parameters. Such a computation complexity is in some sense optimal, as it is the same as the complex- ity needed for computing the sample covariance matrix. Numerical studies show that our algorithm is much more efficient than other state-of-the-art methods when the dimension p is very large. Key words and phrases: ADMM, High dimension, Penalized quadratic loss, Precision matrix. arXiv:1811.04545v2 [stat.CO] 9 Jul 2019 1 Introduction Precision matrices play an important role in statistical learning and data analysis. On the one hand, estimation of the precision matrix is oftentimes required in various sta- tistical analysis. On the other hand, under Gaussian assumptions, the precision matrix has been widely used to study the conditional independence among the random vari- ables.
    [Show full text]
  • Estimation of the Multivariate Normal Precision and Covariance Matrices in a Star-Shape Model*
    Ann. Inst. Statist. Math. Vol. 57, No. 3, 455-484 (2005) Q2005 The Institute of Statistical Mathematics ESTIMATION OF THE MULTIVARIATE NORMAL PRECISION AND COVARIANCE MATRICES IN A STAR-SHAPE MODEL* DONGCHU SUN AND XlAOQIAN SUN Department of Statistics, University of Missouri, Columbia, MO 65211, U.S.A. (Received March 31, 2004; revised September 27, 2004) Abstract. In this paper, we introduce the star-shape models, where the precision matrix ~t (the inverse of the covariance matrix) is structured by the special condi- tional independence. We want to estimate the precision matrix under entropy loss and symmetric loss. We show that the maximal likelihood estimator (MLE) of the precision matrix is biased. Based on the MLE, an unbiased estimate is obtained. We consider a type of Cholesky decomposition of Ft, in the sense that Ft = @~@, where is a lower triangular matrix with positive diagonal elements. A special group G, which is a subgroup of the group consisting all lower triangular matrices, is intro- duced. General forms of equivariant estimates of the covariance matrix and precision matrix are obtained. The invariant Haar measures on G, the reference prior, and the Jeffreys prior of @ are also discussed. We also introduce a class of priors of @, which includes all the priors described above. The posterior properties are discussed and the closed forms of Bayesian estimators are derived under either the entropy loss or the symmetric loss. We also show that the best equivariant estimators with respect to G is the special case of Bayesian estimators. Consequently, the MLE of the preci- sion matrix is inadmissible under either entropy or symmetric loss.
    [Show full text]
  • Iterative Algorithms for Large Stochastic Matrices
    View metadata, citation and similar papers at core.ac.uk brought to you by CORE provided by Elsevier - Publisher Connector Iterative Algorithms for Large Stochastic Matrices Pierre Semal Philips Research Laboratory Au. Albert Einstein 4 B-1348 Louvain-la-Neuve, Belgium Dedicated to Gene Golub, Richard Varga, and David Young Submitted by Robert J. Plemmons ABSTRACT Markov chains have always constituted an efficient tool to model discrete systems. Many performance criteria for discrete systems can be derived from the steady-state probability vector of the associated Markov chain. However, the large size of the state space of the Markov chain often allows this vector to be determined by iterative methods only. Various iterative methods exist, but none can be proved a priori to be the best. In this paper, we propose a practical measure which allows the convergence rate of the various existing methods to be compared. This measure is an approxima- tion of the modulus of the second largest eigenvalue of the iteration matrix and can be determined a priori. The model of a queueing network is used as an example to compare the convergence of several iterative methods and to show the accuracy of the measure. 1. INTRODUCTION Markov chains have always constituted an efficient tool to model discrete systems. If the system can be completely characterized by a state variable, then its behavior is completely described by the matrix Q of transition probabilities between the possible values (states) of that state variable. Most of the performance measures of the system could then be derived from the steady-state vector x of that matrix Q.
    [Show full text]
  • Proximal Approaches for Matrix Optimization Problems: Application to Robust Precision Matrix Estimation?
    Proximal Approaches for Matrix Optimization Problems: Application to Robust Precision Matrix Estimation? A. Benfenatia,∗, E. Chouzenouxb, J.{C. Pesquetb aDipartimento di Scienze e Politiche Ambientali, Universit´adegli studi di Milano, Via Celoria 2, 20133, Milano, Italy bCenter for Visual Computing, INRIA Saclay and CentraleSup´elec, University Paris-Saclay, 9 rue Joliot{Curie, 91190, Gif{sur{Yvette, France Abstract In recent years, there has been a growing interest in mathematical models leading to the minimization, in a symmetric matrix space, of a Bregman divergence coupled with a regularization term. We address problems of this type within a general framework where the regularization term is split into two parts, one being a spectral function while the other is arbitrary. A Douglas{Rachford approach is proposed to address such problems, and a list of proximity operators is provided allowing us to consider various choices for the fit–to–data functional and for the regularization term. Based on our theoretical results, two novel approaches are proposed for the noisy graphical lasso problem, where a covariance or precision matrix has to be statistically estimated in presence of noise. The Douglas{Rachford approach directly applies to the estimation of the covariance matrix. When the precision matrix is sought, we solve a non- convex optimization problem. More precisely, we propose a majorization{minimization approach building a sequence of convex surrogates and solving the inner optimization subproblems via the aforementioned Douglas{Rachford procedure. We establish conditions for the convergence of this iterative scheme. We illustrate the good numerical performance of the proposed approaches with respect to state{of{the{art approaches on synthetic and real-world datasets.
    [Show full text]