# MS Resume

Monday, June 18

MS 1. **Recent advances in matrix functions **

Organizer: Edvin Deadman

University of Manchester, UK

Organizer: Nicholas J. Higham

University of Manchester, UK

Matrix functions are of growing interest in science, engineering and the social sciences, due to the succinct and insightful way they allow problems to be formulated and solutions to be expressed. Many challenges remain in the computation of matrix functions. This minisymposium focuses on some recent advances, including preservation of structure, convergence of iterations, and design of algorithms that exploit modern computer architectures.

11:00–11:25 *Computational issues related to the geometric mean of structured matrices *

Dario A. Bini, Università di Pisa, Italy; Bruno Iannazzo, Università di Perugia, Italy

11:25–11:50 *Efﬁcient, communication-minimizing algorithms for the symmetric eigenvalue decomposition and the singular value decomposition*

Yuji Nakatsukasa, University of Manchester; Nicholas J. Higham, University of Manchester

11:50–12:15 *The Padé approximation and the matrix sign function*

Krystyna Zietak, Wrocław University of Technology

12:15–12:40 *A recursive blocked Schur algorithm for computing the matrix square root*

Edvin Deadman, The University of Manchester / NAG; Nicholas J. Higham, University of Manchester; Rui Ralha, University of Minho, Portugal

Monday, June 18

MS 2. **Methods for Toeplitz matrices and their application**

Organizer: Matthias Bolten

University of Wuppertal, Germany

Many applications, e.g., in image processing, physics or ﬁnance, lead to Toeplitz matrices, i.e. matrices with constant entries on the diagonals. The Toeplitz structure and the associated generating symbol allows for the rigorous analysis of the methods used within the applications, including iterative methods like multigrid methods or methods to compute matrix functions. Many of these results carry over to block matrices with Toeplitz structure, as well. Within this minisymposium new developments regarding methods for Toeplitz systems will be presented. Further, applications for Toeplitz matrices and Toeplitz-block matrices are shown.

11:00–11:25 *Toeplitz operators with matrix-valued symbols and some (unexpected) applications*

Stefano Serra Capizzano, Università dell’Insubria -sede di Como

11:25–11:50 *Fast approximation to the Toeplitz matrix exponential*

Hai-Wei Sun, University of Macau

11:50–12:15 *Matrix algebras sequences can be spectrally equivalent with ill-conditioned Toeplitz ones*

Paris Vassalos, Athens University of Economics and Business; Dimitrios Noutsos, University of Ioannina

12:15–12:40 *Aggregation-based multigrid methods for Toeplitz matrices*

Matthias Bolten, University of Wuppertal; Marco Donatelli, Università dell’Insubria sede di Como; Thomas Huckle, Technische Universität München

Monday, June 18

MS 3. **Matrix factorizations and applications**

Organizer: Michael Tsatsomeros

Washington State Universiy, USA

Organizer: Rafael Cantó

Universitat Politècnica de València, ES

Factorizations of matrices play a crucial and diverse role in applied linear algebra. They typically lead to the development of numerical methods and theoretical understanding of diverse concepts in statistics, optimization, economics and others. In this minisymposium we assemble four talks on different types and applications of matrix factorizations. Particular attention will be paid to bidiagonal factorizations of some classes of matrices, the Cholesky full rank factorization for rank deﬁcient matrices, the SVD decomposition with a number of related applications and a nonnegative matrix factorization.

11:00–11:25 *Classes of matrices with bidiagonal factorization*

Álvaro Barreras, Universidad de Zaragoza; J.M. Peña, Univ. de Zaragoza

11:25–11:50 *Cholesky factorization for singular matrices*

Rafael Cantó, Universitat Politècnica de València; A.M. Urbano, Univ. Politècnica de València; M.J. Peláez, Univ. Católica del Norte

11:50–12:15 *Applications of the singular value decomposition to perturbation theory of eigenvalues of matrix polynomials*

Panayiotis Psarrakos, National Technical University of Athens; N. Papathanasiou, National Technical Univ. of Athens

12:15–12:40 *On reduced rank nonnegative matrix factorization for symmetric nonnegative matrices*

Minerva Catral, Xavier University; L. Han, Univ. of Michigan-Flint; M. Neumann, Univ. of Connecticut; R.J. Plemmons, Wake Forest Univ.

Monday, June 18

MS 4. **Algorithms on manifolds of low-rank matrices and tensors**

Organizer: Bart Vandereycken

École Polytechnique Fédérale de Lausanne, Switzerland

Organizer: Pierre-Antoine Absil

Université catholique de Louvain, Belgium

This minisymposium concerns numerical algorithms on manifolds of tensors and matrices with certain rank structures. While exploiting rank is standard in numerical linear algebra, low-rank algorithms are usually based on techniques that are difﬁcult to analyze. On the other hand, considering the set of ﬁxed rank matrices/tensors as a smooth manifold provides a differential geometric framework for deriving and analyzing such low-rank algorithms. The speakers present recent developments in the ﬁeld of pseudospectra, model reduction and high-dimensional equations. They show how such a differential geometric approach leads to efﬁcient numerical algorithms and point out the advantages compared to more standard approaches.

11:00–11:25 *Low rank dynamics for computing extremal points of real and complex pseudospectra*

Nicola Guglielmi, Università di L’Aquila; Christian Lubich, Universität Tübingen

11:25–11:50 *Parametric model order reduction using stabilized consistent interpolation on matrix manifolds*

David Amsallem, Stanford University; Charbel Farhat, Stanford University

11:50–12:15 *Treatment of high-dimensional problems by low-rank manifolds of tensors*

Thorsten Rohwedder, Technische Universität Berlin

12:15–12:40 *Local convergence of alternating optimization of multivariate functions in the presence of scaling indeterminacies*

André Uschmajew, Technische Universität Berlin

Monday, June 18

MS 5. **Advances in algebraic multigrid -New approaches and applications**

Organizer: Irad Yavneh

Institute of Technology, Israel

Organizer: Eran Treister

Institute of Technology, Israel

Algebraic Multigrid (AMG) methods have long been recognized for their efﬁciency as solvers of sparse systems of equations, mainly such that arise from discretizations of PDEs. However, when it comes to solving more general algebraic systems, some drawbacks are evident in the classical methods. Consequently, a great effort is invested in extending the applicability of AMG methods for the solution of new and more complicated problems. This minisymposium focusses on the development of new AMG algorithms for solving or preconditioning linear systems arising in practical applications. Examples of such applications are found in numerical models used in computer networks, ﬂow simulations, and elasticity.

11:00–11:25 *Algebraic collocation coarse approximation multigrid*

Eran Treister, Institute of Technology; Ran Zemach, Technion; Irad Yavneh, Technion

11:25–11:50 *Energy-minimization interpolation for adaptive algebraic multigrid*

Jacob B. Schroder, Lawrence Livermore National Laboratory; Robert D. Falgout, Lawrence Livermore National Laboratory

11:50–12:15 *Algebraic multigrid (AMG) for complex network calculations*

Geoffrey D. Sanders, Lawrence Livermore National Laboratory; Van Emden Henson, Lawrence Livermore National Laboratory; Panayot S. Vassilevski, Lawrence Livermore National Laboratory

12:15–12:40 *The polynomial of best uniform approximation to 1/x as smoother in two grid methods*

Ludmil T. Zikatanov, The Pennsylvania State University; Johannes Kraus, RICAM, Austria; Panayot S. Vassilevski, Lawrence Livermore National Laboratory

Monday, June 18

MS 6. **Recent advances in fast iterative solvers -Part I of II**

Organizer: Chen Greif

The University of British Columbia, Canada

Organizer: Alison Ramage

University of Strathclyde, UK

This two-session minisymposium concerns the solution of large and sparse linear systems using iterative solvers. We will address a variety of timely themes that pertain to preconditioning, convergence properties of solvers, and relevant applications. Speciﬁc topics include new preconditioned iterative schemes, the analysis of norms and spectral properties for measuring convergence of iterative solvers, and applications in ﬂuid ﬂow and liquid crystal.

11:00–11:25 *Challenges in analysis of Krylov subspace methods*

Zdenek Strakos, Charles University in Prague and Academy of Science of the Czech Republic; Jörg Liesen, Technical University of Berlin

11:25–11:50 *Updating preconditioners for parameterized systems*

Eric de Sturler, Virginia Tech; Sarah Wyatt, Virginia Tech; Serkan Gugercin, Virginia Tech

11:50–12:15 *Efﬁcient preconditioning techniques for two-phase ﬂow simulations*

Maya Neytcheva, Uppsala University; Owe Axelsson, Academy of Sciences of the Czech Republic; Petia Boyanova, Uppsala University; Martin Kronbichler, Uppsala University; Xunxun Wu, Uppsala University

12:15–12:40 *Preconditioners in liquid crystal modelling*

Alison Ramage, University of Strathclyde; Chris Newton, Hewlett-Packard Laboratories

Monday, June 18

MS 7. **Application of statistics to numerical linear algebra algorithms -Part I of II**

Organizer: Marc Baboulin

INRIA Saclay Île-de-France and University Paris-Sud, France

Organizer: Haim Avron

IBM T. J. Watson Research Center, USA

The last several years saw many new algorithms in applied linear algebra employing statistical methods. This increased interest is motivated by the fact that the resulting algorithms are able to outperform deterministic methods while still providing very accurate results. This mini-symposium will address innovative statistical approaches that accelerate signiﬁcantly the solution of either dense or sparse linear systems and also enable us to estimate the conditioning of the solution. The speakers will also present several applications of randomized algorithms in numerical linear algebra including low rank approximations, numerical issues and how these algorithms adapt to large-scale parallel environment.

11:00–11:25 *Fast linear system solvers based on randomization techniques*

Marc Baboulin, Inria Saclay Île-de-France and University Paris-Sud, France; Dulceneia Becker, University of Tennessee, USA; Jack Dongarra, University of Tennessee, USA; Stanimire Tomov, University of Tennessee, USA

11:25–11:50 *Numerical issues in randomized algorithms*

Ilse Ipsen, North Carolina State University, USA

11:50–12:15 *Near-optimal column based matrix reconstruction*

Christos Boutsidis, IBM T. J. Watson Research Center, USA; Petros Drineas, Rensselaer Polytechnic Institute, USA; Malik Magdon-Ismail, Rensselaer Polytechnic Institute, USA

12:15–12:40 *Numerical experiments with statistical condition estimation*

Alan J. Laub, University of California Los Angeles, USA

Monday, June 18

MS 8. **Rational Krylov methods: analysis and applications -Part I of II**

Organizer: Vladimir Druskin

Schlumberger–Doll Research, USA

Organizer: Stefan Güttel

University of Oxford, UK

Rational Krylov spaces are a natural generalization of polynomial Krylov spaces to rational functions. In the last years, rational Krylov methods have proven to be useful tools for a variety of problems, such as the efﬁcient solution of linear and nonlinear eigenvalue problems or matrix equations, in model order reduction, or the computation of matrix functions. This minisymposium brings together leading experts in the analysis and application of these methods, providing a lively forum on this central research topic of numerical linear algebra.

11:00–11:25 *Solving Sylvester equations through rational Galerkin projections*

Bernhard Beckermann, University of Lille

11:25–11:50 *Stability-corrected spectral Lanczos decomposition algorithm for wave propagation in unbounded domains*

Rob Remis, Delft University of Technology; Vladimir Druskin, Schlumberger–Doll Research

11:50–12:15 *Generalized rational Krylov decompositions*

Stefan Güttel, University of Oxford

12:15–12:40 *Interpolatory model reduction strategies for nonlinear parametric inversion*

Serkan Gugercin, Virginia Tech.; Christopher A. Beattie, Virginia Tech.; Saifon Chaturantabut, Virginia Tech.; Eric de Sturler, Virginia Tech.; Misha E. Kilmer, Tufts University

Monday, June 18

MS 9. **New trends in tridiagonal matrices -Part I of II**

Organizer: Natália Bebiano

University of Coimbra, Portugal

Tridiagonal matrices emerge in plenty of applications in science and engineering. They are used for solving a variety of problems in disparate contexts. Beyond their several applications seldom discussed, the methods, techniques, and theoretical framework used in this research ﬁeld make it very interesting and challenging. In this forum of discussion, recent developments in the area are focused and new approaches and perspectives are searched.

11:00–11:25 *Direct and inverse problems on pseudo-Jacobi matrices*

Natália Bebiano, University of Coimbra, Portugal; Susana Furtado, University of Porto, Portugal; J. da Providência, University of Coimbra, Portugal

11:25–11:50 *Schwartz’s matrices and generalized Hurwitz polynomials*

Mikhail Tyaglov, Technische Universität Berlin, Germany

11:50–12:15 *On the Moore-Penrose inverse of singular, symmetric and periodic Jacobi M-matrices*

Andrés M. Encinas, Enrique Bendito, Ángeles Carmona and Margarida Mitjana, Universitat Politècnica de Catalunya, Spain

12:15–12:40 *The commutant of the tridiagonal pattern*

Charles R. Johnson, College of William and Mary, Williamsburg, USA

Monday, June 18

MS 10. **Numerical algorithms for switching systems: from theory to applications**

Organizer: Nicola Guglielmi

Universita degli Studi dell’ Aquila, Italy

Organizer: Raphaël Jungers

FNRS and UCLouvain, Belgium

In the past decade much progress has been achieved in the understanding of switching linear systems, and the development of algorithms to control/optimize these systems has received a strong impulse. On the one hand many fundamental theoretical issues have been investigated and partly settled. On the other hand, many applications have emerged, which can be cast as switching systems. These applications touch upon many different ﬁelds of engineering, from decentralized control to drug treatment optimization. The goal of the workshop is to present these two facets from a linear algebra perspective, and investigate the gap between theory and practice.

15:00–15:25 *Observer design for hybrid systems*

M. D. Di Benedetto, University of L’Aquila

15:25–15:50 *About polynomial instability for linear switched systems*

P. Mason, CNRS; Y. Chitour, University of Paris Sud; M. Sigalotti, INRIA Saclay

15:50–16:15 *Stability and stabilization of positive switched systems: state of the art and open problems*

M.E. Valcher, University of Padova; E. Fornasini, University of Padova

16:15–16:40 *The joint spectral radius for semigroups generated by switched differential algebraic equations*

F. Wirth, University of Würzburg; S. Trenn, Technical University of Kaiserslautern

Monday, June 18

MS 11. **Recent advances in fast iterative solvers -Part II of II**

Organizer: Chen Greif

The University of British Columbia, Canada

Organizer: Alison Ramage

University of Strathclyde, UK

This two-session minisymposium concerns the solution of large and sparse linear systems using iterative solvers. We will address a variety of timely themes that pertain to preconditioning, convergence properties of solvers, and relevant applications. Speciﬁc topics include new preconditioned iterative schemes, the analysis of norms and spectral properties for measuring convergence of iterative solvers, and applications in ﬂuid ﬂow and liquid crystal.

15:00–15:25 *Combination preconditioning of saddle-point systems for positive deﬁniteness*

Andy Wathen, Oxford University; Jennifer Pestana, Oxford University

15:25–15:50 *Preconditioned iterative methods for nonsymmetric matrices and nonstandard inner products*

Jennifer Pestana, Oxford University; Andy Wathen, Oxford University

15:50–16:15 *Multi-preconditioned GMRES*

Tyrone Rees, Rutherford Appleton Laboratory; Chen Greif, The University of British Columbia; Daniel B. Szyld, Temple University

16:15–16:40 *Bounds on the eigenvalues of indeﬁnite matrices arising from interior-point methods*

Chen Greif, The University of British Columbia; Erin Moulding, The University of British Columbia; Dominique Orban, Ecole Polytechnique de Montreal

Monday, June 18

MS 12. **Application of statistics to numerical linear algebra algorithms -Part II of II**

Organizer: Marc Baboulin

INRIA Saclay Île-de-France and University Paris-Sud, France

Organizer: Haim Avron

IBM T. J. Watson Research Center, USA

The last several years saw many new algorithms in applied linear algebra employing statistical methods. This increased interest is motivated by the fact that the resulting algorithms are able to outperform deterministic methods while still providing very accurate results. This mini-symposium will address innovative statistical approaches that accelerate signiﬁcantly the solution of either dense or sparse linear systems and also enable us to estimate the conditioning of the solution. The speakers will also present several applications of randomized algorithms in numerical linear algebra including low rank approximations, numerical issues and how these algorithms adapt to large-scale parallel environment.

15:00–15:25 *Spectral graph theory, sampling matrix sums, and near-optimal SDD solvers*

Ioannis Koutis, University of Puerto Rico; Gary Miller, Carnegie Mellon University, USA; Richard Peng, Carnegie Mellon University, USA

15:25–15:50 *Implementation of a randomization algorithm for dense linear algebra libraries*

Dulceneia Becker, University of Tennessee, USA; Marc Baboulin, Inria Saclay Île-de-France and University Paris-Sud, France; Jack Dongarra, University of Tennessee, USA

15:50–16:15 *Implementing randomized matrix algorithms in large-scale parallel environments*

Michael W. Mahoney, Stanford University, USA

16:15–16:40 *Random sampling preconditioners*

Haim Avron, IBM T. J. Watson Research Center, USA; Sivan Toledo, Tel-Aviv University, Israel

Monday, June 18

MS 13. **Rational Krylov methods: analysis and applications -Part II of II**

Organizer: Vladimir Druskin

Schlumberger–Doll Research, USA

Organizer: Stefan Güttel

University of Oxford, UK

Rational Krylov spaces are a natural generalization of polynomial Krylov spaces to rational functions. In the last years, rational Krylov methods have proven to be useful tools for a variety of problems, such as the efﬁcient solution of linear and nonlinear eigenvalue problems or matrix equations, in model order reduction, or the computation of matrix functions. This minisymposium brings together leading experts in the analysis and application of these methods, providing a lively forum on this central research topic of numerical linear algebra.

15:00–15:25 *Rational Krylov methods for nonlinear matrix problems*

Karl Meerbergen, KU Leuven; Roel Van Beeumen, KU Leuven; Wim Michiels, KU Leuven

15:25–15:50 *Block Gauss and anti-Gauss quadrature rules with application to networks*

Lothar Reichel, Kent State University; David Martin, Kent State University

15:50–16:15 *On optimality of rational Krylov based low-rank approximations of large-scale matrix equations*

Tobias Breiten, Max Planck Institute for Dynamics of Complex Technical Systems; Peter Benner, Max Planck Institute for Dynamics of Complex Technical Systems

16:15–16:40 *Inverse problems for large-scale dynamical systems in the H2-optimal model reduction framework*

Mikhail Zaslavsky, Schlumberger–Doll Research; Vladimir Druskin, Schlumberger–Doll Research; Valeria Simoncini, University of Bologna

Monday, June 18

MS 14. **New trends in tridiagonal matrices -Part II of II**

Organizer: Carlos Fonseca

University of Coimbra, Portugal

Tridiagonal matrices emerge in plenty of applications in science and engineering. They are used for solving a variety of problems in disparate contexts. Beyond their several applications seldom discussed, the methods, techniques, and theoretical framework used in this research ﬁeld make it very interesting and challenging. In this forum of discussion, recent developments in the area are focused and new approaches and perspectives are searched.

15:00–15:25 *On generalized Jacobi matrices which are symmetric in Krein spaces*

Maxim Derevyagin, Technische Universität Berlin, Germany

15:25–15:50 *On the characteristic function for Jacobi matrices*

Pavel Stovícek, Czech Technical University, Czech Republic

15:50–16:15 *Tridiagonal matrices in comb ﬁlters*

Jesús Gutiérrez-Gutiérrez, Universidad de Navarra, Spain

16:15–16:40 *The nullity theorem: forecasting structures in the inverses of sparse matrices*

Raf Vandebril, K.U. Leuven, Belgium

Monday, June 18

MS 15. **Application of compressed sensing in Bio-Medicine**

Organizer: Amir Niknejad

The College of Mount Saint Vincent, USA

The past decade have witnessed burgeoning research activity in the area of compressed sensing and sparse approximation. This minisymposium is devoted to the mathematical aspects of sparsity in undetermined linear systems and its applications in bio-medicine. It will bring together Scientists who use compressive sensing in their respected ﬁelds. The focus is on problems arising in molecular biology and bio-medicine. Most application is related to the processing of biological data, Drug Discovery, and neuro imaging. This minisymposium is intended as a workshop for strengthening communication between computational biologists and the linear algebra community for fostering collaborations.

15:00–15:25 *Evaluation of compressed sensing impact in cardiac signals processing and transmission*

Eduardo Pinheiro, Instituto de Telecomunicaes Instituto Superior Tcnico, Lisboa, Portugal

15:25–15:50 *Compressive sensing in drug discovery*

Marcus Weber, Zuse Institute Berlin (ZIB), Germany

15:50–16:15 *Reconstruction of bacterial communities using sparse representation*

Or Zuk, Broad Institute, Cambridge, USA

16:15–16:40 *Sensing genome via factorization*

Amir Niknejad, College of Mount Saint Vincent,New York, USA

Monday, June 18

MS 16. **Preconditioning of non-normal linear systems arising in scattering problems**

Organizer: Kees Vuik and Neil Budko

Delft University of Technology, The Netherlands

The scattering of waves on inhomogeneous objects and many other important problems are described by large linear systems with non-normal matrices. A few iterative methods that can cope with such systems often suffer from an extremely slow convergence. In recent years there has been some progress in understanding the reasons behind this behavior and several efﬁcient preconditioners have been proposed ranging from the deﬂation of the largestmagnitude eigenvalues, to regularization, the approximate factorization of the inverse, and shifted Laplacians. In our minisymposium we review the application of these methods to differential and integral equations arising in scattering problems.

15:00–15:25 *Approximate deﬂation preconditioning methods for penetrable scattering problems*

Josef Sifuentes, New York University; Mark Embree, Rice University

15:25–15:50 *Direct approximate factoring of the inverse*

Marko Huhtanen, University of Oulu; Mikko Byckling, CERFACS

15:50–16:15 *Regularization of singular integral operators as a preconditioning strategy*

Neil Budko, Delft University of Technology; Grigorios Zouros, National Technical University of Athens

16:15–16:40 *High-order shifted Laplace preconditioners for wave equations*

Xavier Antoine, University of Lorraine; Christophe Geuzaine, University of Liège; Ibrahim Zangré, University of Lorraine

Monday, June 18

MS 17. **Markov chains**

Organizer: Jeffrey J. Hunter

Auckland University of Technology, New Zealand

Organizer: Stephen J. Kirkland

National University of Ireland, Ireland

The mini-symposium will feature recent research activity in the area of Markov chains, with a focus on some key properties and applications. Interpretations of Kemeny’s constant, the interaction between random walks on graphs and electrical networks, and matrix-based approaches, all offer signiﬁcant opportunities for new advances. These approaches will be considered in several contexts, including hitting and mixing times, ergodicity, Laplacians, Hamiltonian cycles, generalized inverses, perturbation and sensitivity analysis, computational procedures, the column sums of a transition matrix, and different matrix representations. The mini-symposium is expected to encourage further dialogue between the theory and applications of Markov chains.

15:00–15:25 *Markov chain properties in terms of column sums of the transition matrix*

Jeffrey J. Hunter, Auckland University of Technology, Auckland, New Zealand

15:25–15:50 *Hamiltonian cycle problem and Markov chains*

Jerzy Filar, Flinders University, Bedford Park, SA, Australia

15:50–16:15 *Inequalities for functions of transition matrices*

Iddo Ben-Ari, University of Connecticut, Storrs, CT, United States of America

16:15–16:40 *Compartmental systems and computation of their stationary probability vectors*

Ivo Marek, Czech Institute of Technology, Praha, Czech Republic

Tuesday, June 19

MS 18. **Preconditioning for PDE-constrained optimization -Part I of II**

Organizer: Martin Stoll

Computational Methods in Systems and Control Theory, Max Planck Institute for Dynamics of Complex Technical Systems, Germany

Organizer: Andy Wathen

Numerical Analysis Group, United Kingdom

The solution of optimization problems with constraints given by partial differential equations is a challenging problem for numerical analysts. Whether the constraint is linear or nonlinear, at the heart of the optimization procedure lies the solution of linear systems that are often very large, sparse and structured. Our minisymposium is aimed at providing insights into recent developments with respect to Krylov subspace methods that in conjunction with efﬁcient preconditioning techniques provide a robust and ﬂexible framework for solving PDE-constrained optimization problems.

11:00–11:25 *Structural spectral properties of symmetric saddle point problems*

Valeria Simoncini, Università di Bologna, Italy, Wolfgang Krendl and Walter Zulehner, Johannes Kepler University, Austria

11:25–11:50 *Preconditioned iterative methods for Stokes and Navier-Stokes control problems*

John Pearson, University of Oxford, UK

11:50–12:15 *Preconditioners for elliptic optimal control problems with inequality constraints*

Walter Zulehner and Markus Kollmann, Johannes Kepler University, Austria

12:15–12:40 *Nearly optimal block preconditioners for block two-by-two linear systems*

Zhong-Zhi Bai, Chinese Academy of Sciences, China

Tuesday, June 19

MS 19. **Matrices and graphs -Part I of II**

Organizer: Leslie Hogben

Iowa State University, USA

Organizer: Stephen Kirkland

National University of Ireland Maynooth, Ireland

Interactions between matrices and graphs now play a vital role in both matrix theory and graph theory, and have applications to a variety of ﬁelds. This mini-symposium will present recent results related to the synergistic relationship between matrices and graphs, including properties of matrices having a nonzero pattern described by a given graph, and information about a graph provided by speciﬁc matrices associated with the graph. Applications to control of quantum systems and other areas will be presented.

11:00–11:25 *(0, 1) matrices and the analysis of social networks*

Steve Kirkland, National University of Ireland Maynooth

11:25–11:50 *Necessary and sufﬁcient conditions for a Hamiltonian graph*

Irene Sciriha, University of Malta; Domingos Moreira Cardoso, Univ. de Aveiro

11:50–12:15 *On the eigenvalues of symmetric matrices associated with graphs*

Miriam Farber, Technion

12:15–12:40 *An extension of the polytope of doubly stochastic matrices*

Richard Brualdi, University of Wisconsin-Madison; Geir Dahl, University of Oslo

Tuesday, June 19

MS 20. **Tensor based methods for high dimensional problems in scientiﬁc computing -Part I of II**

Organizer: A. Falcó

Universidad de Alicante, Spain

Organizer: A. Nouy

LUNAM Université, Ecole Centrale Nantes, France

The use of tensor product approximations is receiving a growing interest in numerical analysis for the solution of problems deﬁned in high-dimensional tensor spaces, such as PDEs arising in stochastic calculus (e.g. the Fokker-Planck equation), variational problems, approximation theory, stochastic parametric PDEs arising in uncertainty quantiﬁcation, and quantum chemistry. This minisymposium focuses on recent advances in this topic. In particular, the scope of this proposal includes: (a) High dimensional PDEs (deterministic or stochastic) or other operator equations in tensor format. (b) Iterative methods using tensor approximations. Preconditioning issues, convergence analysis. (c) Alternative deﬁnitions and algorithms for tensor decompositions. (d) Functional Analysis approach of tensor methods. (e) Applications of tensor methods in real-life problems.

11:00–11:25 *Optimal a priori tensor decomposition for the solution of high dimensional problems*

Anthony Nouy, LUNAM Université, Ecole Centrale Nantes

11:25–11:50 *Application of the Proper Generalized Decomposition (PGD) to 3D cracked plates and estimation of the discretization error *

Eugenio Giner, CITV, UPV

11:50–12:15 *A tensor calculus approach for Bézier shape deformation*

Lucía Hilario, U. CEU Cardenal Herrera

12:15–12:40 *Tensor approximation methods for parameter identiﬁcation*

Hermann G. Matthies, Alexander Litvinenko, Bojana Rosíc and Oliver Pajonk, Institute of Scientiﬁc Computing, TU Braunschweig

Tuesday, June 19

MS 21. **Reducing communication in linear algebra -Part I of II**

Organizer: Aydın Buluç

Lawrence Berkeley National Laboratory

Organizer: Oded Schwartz

University of California, Berkeley

Performance scaling of linear algebra kernels is limited by the cost of data movement between memory hierarchy levels and between processors in a parallel setting. Communication efﬁcient linear algebra kernels will help scientiﬁc computing reach exascale, and accelerate non-numerical applications that rely on linear algebra such as machine learning, data mining, and graph analysis. This minisymposium discusses new dense and sparse linear algebra algorithms that move asymptotically less data, lower bounds on the amount of communication needed for various problems, and practical implementations that outperform conventional codes by reducing communication.

11:00–11:25 *Communication-optimal parallel algorithm for Strassen’s matrix multiplication*

Oded Schwartz, Grey Ballard, James Demmel, Olga Holtz and Benjamin Lipshitz, UC Berkeley

11:25–11:50 *A communication-avoiding symmetric-indeﬁnite factorization*

Sivan Toledo, Tel-Aviv University; Grey Ballard and James Demmel, UC Berkeley; Alex Druinsky and Inon Peled, Tel-Aviv University; Oded Schwartz, UC Berkeley

11:50–12:15 *LU factorization with panel rank revealing pivoting and its communication avoiding version*

Amal Khabou, INRIA Saclay -Île de France; James Demmel, UC Berkeley; Laura Grigori, INRIA Saclay -Ile de France; Ming Gu, UC Berkeley

12:15–12:40 *2.5D Algorithms for parallel dense linear algebra*

Edgar Solomonik and James Demmel, UC Berkeley

Tuesday, June 19

MS 22. **Linear algebra for inverse problems -Part I of II**

Organizer: L. Reichel

Kent State University, USA

Organizer: H. Sadok

LMPA, Université du Littoral, France

Many problems in science and engineering require the determination of the cause of an observed effect. These problems often can be formulated as Fredholm integral equations of the ﬁrst kind with a smooth kernel. Their discretization gives rise to linear systems of equations with a matrix whose singular values “cluster” at the origin and whose right-hand side is contaminated by error. The development and analysis of solution techniques for these kinds of problems is an active area of research, and much of this work is based on linear algebra. This minisymposium brings together leading experts on the development and analysis of solution methods for inverse problems. The presentations provide an overview of recent developments with an emphasis on the role of linear algebra.

11:00–11:25 *Block-extrapolation methods for linear matrix ill-posed problem*

K. Jbilou, Université du Littoral

11:25–11:50 *Convergence properties of the GMRES and RRGMES methods for ill-posed problems*

H. Sadok, Université du Littoral

11:50–12:15 *Inverse problems for regularization matrices*

Silvia Noschese, SAPIENZA Università di Roma; Lothar Reichel, Kent State University

12:15–12:40 *Meshless regularization for the numerical computation of the solution of steady Burgers-type equations*

A. Bouhamidi, Université du Littoral; M. Hached, Université du Littoral; K. Jbilou, Université du Littoral

Tuesday, June 19

MS 23. **Modern matrix methods for large scale data and networks**

Organizer: David F. Gleich

Purdue University, USA

Every few years, the new applications for matrix methods arise and challenge existing paradigms. The talks in this mini-symposium sample some of the research that has arisen out of new applications in large scale machine learning, network problems, and data analysis. Much of the research presented at this mini-symposium will have an interesting twist on a classical matrix problem – linear systems, least squares, or eigenvalues – that better ﬁts the current problems.

11:00–11:25 *Nonlinear eigenproblems in data analysis and graph partitioning*

Matthias Hein, Saarland University

11:25–11:50 *LSRN: a parallel iterative solver for strongly over-or under-determined systems*

Xiangrui Meng, M. A. Saunders and M. W. Mahoney, Stanford University

11:50–12:15 *Solving large dense linear systems with covariance matrices*

Jie Chen, Argonne National Laboratory

12:15–12:40 *Fast coordinate descent methods with variable selection for non-negative matrix factorization*

Inderjit S. Dhillon and Cho-Jui Hsieh, The University of Texas at Austin

Tuesday, June 19

MS 24. **Novel and synergetic algorithms for multicore and multinode architecture**

Organizer: Olaf Schenk

University of Lugano, Switzerland

Organizer: Ping Tak Peter Tang

Intel Corporation

The advent of parallel computers from multiple CPUs to, more recently, multiple processor cores within a single CPU has continued to spur innovative linear algebra algorithms. This minisymposium aims to present a number of works that not only exhibit new algorithmic ideas suitable for these modern computer architecture, but also present interesting synergies among themselves in several levels. Some can act as a plug-and-play component to enable others; the enabled algorithms can produce components to enable the enablers in turn; and special algorithmic developments were undertaken to enrich this ecosystem.

11:00–11:25 *Novel and synergetic linear algebra algorithms on multicore and multinode architecture*

Ping Tak Peter Tang, Intel Corporation

11:25–11:50 *PSPIKE – A hybrid sparse linear system solver*

Olaf Schenk, University of Lugano, Switzerland; Ahmed Sameh, Purdue University

11:50–12:15 *Eigensolver Based Reordering and Parallel TraceMIN*

Murat Manguogluo, Middle East Technical University; Faisal Saied and Ahmed Sameh, Purdue University

12:15–12:40 *FEAST – A density matrix based eigensolver*

Eric Polizzi, University of Massachusetts

Tuesday, June 19

MS 25. **Direction preserving and ﬁltering methods for solving sparse linear systems**

Organizer: Laura Grigori

University Paris 11, France

Organizer: Frederic Nataf

Paris 6 University, France

This minisymposium reviews several results obtained in the recent years in developing direction preserving and ﬁltering methods for solving sparse linear systems of equations. These methods have been studied in different contexts. For example, preconditioners that are identical with the input matrix given a set of vectors are able to deal efﬁciently with low frequency modes. For multigrid methods, the ﬁltering allows to preserve in the coarse space several directions of interest. The talks discuss this approach for two level domain decomposition methods, multigrid methods, and approached block factorizations.

11:00–11:25 *Algebraic two-level domain decomposition methods*

Frederic Nataf, Paris 6 University, France; Pascal Have, IFP, France; Roland Masson, University of Nice, France; Mikolaj Szydlarski and Tao Zhao, University of Paris 6, France

11:25–11:50 *Filtering solvers*

G. Wittum, University of Frankfurt, Germany

11:50–12:15 *Bootstrap algebraic multigrid*

Karsten Kahl, University of Wuppertal, Germany; James Brannick, Pennsylvania State University, USA

12:15–12:40 *Block ﬁltering decomposition*

Laura Grigori, INRIA Saclay, University of Paris 11, France; Frederic Nataf, University of Paris 6, France; Riadh Fezzanni, INRIA Saclay, University of Paris 11, France;

Tuesday, June 19

MS 26. **Advances in Krylov subspace methods**

Organizer: Sou-Cheng Choi

The University of Chicago, US

Krylov subspace methods have a long illustrious history in numerical linear algebra. Acronyms like BiCG-Stab, CG, GMRES, LSQR, MINRES, QMR, etc, have become part of the standard vocabulary of every numerical analyst. It is somewhat surprising that major advances are still being made to a subject so classical. This minisymposium brings together researchers who have made recent major breakthroughs in the development of Krylov subspace methods—new algorithms that ﬁll existing gaps, better convergence and stability analyses, and novel adaptations for efﬁciency under alternative measures of computational costs (such as communication complexity).

11:00–11:25 *The new challenges to Krylov subspace methods*

Yousef Saad, The University of Minnesota, USA

11:25–11:50 *Random shadow vectors in IDR(s): an explanation of its GMRES-like convergence*

Peter Sonneveld, The Delft University of Technology

11:50–12:15 *Truncated and inexact Krylov subspace methods for parabolic control problems*

Daniel Szyld, The Temple University, USA ; X. Du, Alfred Univ., USA ; M. Sarkis, Worcester Polytechnic Inst., USA ; C. Schaerer, National Univ. of Asuncion, Paraguay

12:15–12:40 *Convergence of iterative solution algorithms for least-squares problems*

David Titley-Peloquin, The University of Oxford; S. Gratton, ENSEEIHT; P. Jiranek, CERFACS

Tuesday, June 19

MS 27. **Preconditioning forPDE-constrained optimization -Part II of II**

Organizer: Martin Stoll

Computational Methods in Systems and Control Theory Max Planck Institute for Dynamics of Complex Technical Systems, Germany

Organizer: Andy Wathen

Mathematical Institute, United Kingdom

The solution of optimization problems with constraints given by partial differential equations is a challenging problem for numerical analysts. Whether the constraint is linear or nonlinear, at the heart of the optimization procedure lies the solution of linear systems that are often very large, sparse and structured. Our minisymposium is aimed at providing insights into recent developments with respect to Krylov subspace methods that in conjunction with efﬁcient preconditioning techniques provide a robust and ﬂexible framework for solving PDE-constrained optimization problems.

15:00–15:25 *On linear systems arising in trust-region methods*

Susann Mach and Roland Herzog, University of Technology Chemnitz, Germany

15:25–15:50 *Preconditioning for PDE-constrained optimization using proper orthogonal decomposition*

Ekkehard Sachs and Xuancan Ye, Universitt Trier, Germany

15:50–16:15 *Preconditioning for Allen-Cahn problems with non-local constraints*

Luise Blank, University of Regensburg; Martin Stoll, Max Planck Institute for Dynamics of Complex Technical Systems, Germany; Lavinia Sarbu

16:15–16:40 *A one-shot approach to time-dependent PDE control*

Martin Stoll, Max Planck Institute for Dynamics of Complex Technical Systems, Germany, Andy Wathen and John Pearson; University of Oxford, UK

Tuesday, June 19

MS 28. **Matrices and graphs -Part II of II**

Organizer: Leslie Hogben

Iowa State University, USA

Organizer: Stephen Kirkland

National University of Ireland Maynooth, Ireland

Interactions between matrices and graphs now play a vital role in both matrix theory and graph theory, and have applications to a variety of ﬁelds. This mini-symposium will present recent results related to the synergistic relationship between matrices and graphs, including properties of matrices having a nonzero pattern described by a given graph, and information about a graph provided by speciﬁc matrices associated with the graph. Applications to control of quantum systems and other areas will be presented.

15:00–15:25 *Parameters related to maximum nullity, zero forcing number, and tree-width of a graph*

Leslie Hogben, Iowa State University and American Institute of Mathematics; and others (listed with abstract)

15:25–15:50 *Colin de Verdière numbers of chordal and split graphs*

Felix Goldberg, National University of Ireland Maynooth

15:50–16:15 *Kochen-Specker sets and the rank-1 quantum chromatic number*

Simone Severini, University College London

16:15–16:40 *On the null vectors of graphs*

Shaun Fallat, University of Regina

Tuesday, June 19

MS 29. **Tensor based methods for high dimensional problems in scientiﬁc computing -Part II of II**

Organizer: A. Falcó

Universidad de Alicante, Spain

Organizer: A. Nouy

LUNAM Université, Ecole Centrale Nantes, France

The use of tensor product approximations is receiving a growing interest in numerical analysis for the solution of problems deﬁned in high-dimensional tensor spaces, such as PDEs arising in stochastic calculus (e.g. the Fokker-Planck equation), variational problems, approximation theory, stochastic parametric PDEs arising in uncertainty quantiﬁcation, and quantum chemistry. This minisymposium focuses on recent advances in this topic. In particular, the scope of this proposal includes: (a) High dimensional PDEs (deterministic or stochastic) or other operator equations in tensor format. (b) Iterative methods using tensor approximations. Preconditioning issues, convergence analysis. (c) Alternative deﬁnitions and algorithms for tensor decompositions. (d) Functional Analysis approach of tensor methods. (e) Applications of tensor methods in real-life problems.

15:00–15:25 *A greedy algorithm for the convergence of a Laplacian operators in the blind deconvolution problem*

Pantaleón David Romero Sánchez, Universidad CEU-Cardenal Herrera, Spain

15:25–15:50 *Algorithms for approximate inverse of operators for preconditioning systems of equations in tensor format*

Loic Giraldi LUNAM Université; Anthony Nouy and Gregory Legrain, Université de Nante, France

15:50–16:15 *Geometric structures in tensor representations*

Antonio Falcó, Universidad de Alicante, Spain

Tuesday, June 19

MS 30. **Reducing communication in linear algebra -Part II of II**

Organizer: Aydın Buluç

Lawrence Berkeley National Laboratory, USA

Organizer: Oded Schwartz

University of California, USA

Performance scaling of linear algebra kernels is limited by the cost of data movement between memory hierarchy levels and between processors in a parallel setting. Communication efﬁcient linear algebra kernels will help scientiﬁc computing reach exascale, and accelerate non-numerical applications that rely on linear algebra such as machine learning, data mining, and graph analysis. This minisymposium discusses new dense and sparse linear algebra algorithms that move asymptotically less data, lower bounds on the amount of communication needed for various problems, and practical implementations that outperform conventional codes by reducing communication.

15:00–15:25 *Communication-avoiding sparse matrix-matrix multiplication*

Aydın Buluç, LBNL; Grey Ballard, UC Berkeley; James Demmel, UC Berkeley; Laura Grigori, INRIA Saclay -Île de France; Oded Schwartz, UC Berkeley

15:25–15:50 *Improving the stability of communication-avoiding Krylov subspace methods*

Erin Carson, Nicholas Knight and James Demmel, UC Berkeley, USA

15:50–16:15 *Hiding global synchronization latencies in Krylov methods for systems of linear equations*

Pieter Ghysels and Wim Vanroose University of Antwerp, BE

16:15–16:40 *Avoiding communication with hierarchical matrices*

Nicholas Knight, Erin Carson and James Demmel, UC Berkeley, USA

Tuesday, June 19

MS 31. **Linear algebra for inverse problems -Part II of II**

Organizer: L. Reichel

Kent State University, USA

Organizer: H. Sadok

LMPA, Université du Littoral, France

Many problems in science and engineering require the determination of the cause of an observed effect. These problems often can be formulated as Fredholm integral equations of the ﬁrst kind with a smooth kernel. Their discretization gives rise to linear systems of equations with a matrix whose singular values “cluster” at the origin and whose right-hand side is contaminated by error. The development and analysis of solution techniques for these kinds of problems is an active area of research, and much of this work is based on linear algebra. This minisymposium brings together leading experts on the development and analysis of solution methods for inverse problems. The presentations provide an overview of recent developments with an emphasis on the role of linear algebra.

15:00–15:25 *Implicit ﬁltering methods for inverse problems*

J. G. Nagy, Emory University; A. Cornelio, University of Modena and Reggio Emilia; E. L. Piccolomini, University of Bologna

15:25–15:50 *Iterative reconstruction methods for adaptive optics*

R. Ramlau, Kepler University

15:50–16:15 *Approximated nonstationary iterated Tikhonov with application to image deblurring*

M. Donatelli, University of Insubria; M. Hanke, University of Mainz

16:15–16:40 *On the Richardson-Lucy method for image restoration*

F. Sgallari, University of Bologna; M. K. Khan, Kent State University; S. Morigi, University of Bologna; L. Reichel, Kent State University

Tuesday, June 19

MS 32. **Orderings in sparse matrix computation**

Organizer: Iain S. Duff

Rutherford Appleton Laboratory, UK

Organizer: Esmond G. Ng

Lawrence Berkeley National Laboratory, USA

Ordering in sparse matrix computation refers to the problem of permuting the rows and columns of a sparse matrix to achieve certain objectives. It is an integral step in sparse matrix computation. For example, it is desirable to ﬁnd permutations so that the number of nonzeros created or the number of operations is small in a sparse matrix factorization. In iterative methods, orderings are important in reducing communication in parallel implementations. In this minisymposium, we feature some recent work on sparse matrix orderings. Some of the talks will focus on algorithmic development, while others will investigate the theoretical aspect.

15:00–15:25 *Orderings and solvers for “non-uniform sparse matrices”*

Edmond Chow and Oguz Kaya, Georgia Institute of Technology, USA

15:25–15:50 *On hypergraph partitioning based ordering methods for sparse matrix factorization*

Bora Uçar, CNRS and ENS Lyon; Iain S. Duff, Rutherford Appleton Laboratory; Johannes Langguth, ENS Lyon

15:50–16:15 *Orderings Governed by numerical factorization*

Iain S. Duff and Mario Arioli, Rutherford Appleton Laboratory

16:15–16:40 *Reordering sparse Cholesky factorization: minimum ﬁll vs. minimum FLOP count*

Robert Luce, Technical University of Berlin; Esmond G. Ng, Lawrence Berkeley National Laboratory

Tuesday, June 19

MS 33. **Moving from multicore to manycore in applied linear algebra**

Organizer: Jens Saak

Max Planck Institute for Dyamics of Complex Technical Systems, Germany

Organizer: Alfredo Remón

Universidad Jaume I, Spain

During the last years, the evolution of graphics processors (GPUs) has extended their use to many scientiﬁc and engineering applications. In particular, their highly parallel architecture makes them suitable for vector operations and especially appealing for linear algebra (LA) applications. To leverage their potential, it is necessary to rethink numerical LA algorithms and codes. The impact of GPUs in LA has motivated the design of a number of recent LA libraries: CUBLAS, CUFFT, FLAME, MAGMA, … To illustrate these ongoing developments, we have chosen a handful of researchers that contributed in moving towards manycore computing during the recent years.

15:00–15:25 *Parallel preconditioners and multigrid methods for sparse systems on GPUs*

Jan-Philipp Weiss, Dimitar Lukarski and Vincent Heuveline, Karlsruhe Institute of Technology (KIT), Germany

15:25–15:50 *Towards a GPU-accelerated direct sparse solver*

Pablo Ezzatti and Alejandro Zinemanas, Univ. de la República, Uruguay

15:50–16:15 *Unleashing the power of multicore DSPs for matrix computations. The FLAME approach*

Francisco D. Igual, University Jaume I, Spain; Murtaza Ali, Texas Instruments, USA; Robert A. van de Geijn, The University of Texas at Austin, USA

16:15–16:40 *High-performance genome studies*

Lucas Beyer, Diego Fabregat-Traver and Paolo Bientinesi, Aachen Institute for Advanced Study in Computational Engineering Science (AICES), Germany

Tuesday, June 19

MS 34. **Least squares methods and applications**

Organizer: Sanzheng Qiao

McMaster University1, Canada

Organizer: Yimin Wei

Fudan University, P.R. China

This minisymposium provides an overview of recent research in a wide range of topics in linear and nonlinear least squares problems, including block algorithms, applications in wireless communications, perturbation analysis, and preconditioning techniques

11:00–11:25 *Block Gram–Schmidt algorithms with reorthogonalization*

Jesse Barlow, Penn State University, USA

11:25–11:50 *A numerical method for a mixed discrete bilinear least squares problem*

Xiao-Wen Chang, McGill University, Canada

11:50–12:15 *On condition numbers for constrained linear least squares problems*

Huaian Diao, Northeast Normal University, P.R. China

12:15–12:40 *SOR inner-iteration GMRES for underdetermined least squares problems*

Keiichi Morikuni and Ken Hayami, The Graduate University for Advanced Studies, Japan

Wednesday, June 20

MS 35. **Nonlinear eigenvalue problems**

Organizer: K. Meerbergen

K.U. Leuven, Belgium

Organizer: W. Michiels

K.U. Leuven, Belgium

Organizer: C. Lecomte

University of Southampton, UK

The interest in nonlinear eigenvalue problems has increased signiﬁcantly over the last few years. Recent contributions have focused on methods derived from residual inverse iteration and nonlinear Rayleigh quotient iteration, as well as Krylov methods for linear inﬁnite dimensional problems and linearizations of polynomial eigenvalue problems. Those ideas show up as building blocks for solving non-linear non-polynomial eigenvalue problems.

11:00–11:25 *Computable error bounds for nonlinear eigenvalue problems allowing for a minimax characterization*

Heinrich Voss, Hamburg University of Technology, Germany; Kemal Yildiztekin, Hamburg University of Technology, Germany

11:25–11:50 *A restarting technique for the inﬁnite Arnoldi method*

Elias Jarlebring, KTH -Royal institute of technology, Sweden; Karl Meerbergen, K.U.Leuven, Belgium; Wim Michiels, K.U.Leuven, Belgium

11:50–12:15 *Robust successive computation of eigenpairs for nonlinear eigenvalue problems*

Cedric Effenberger, EPF Lausanne, Switzerland

12:15–12:40 *Triangularization of matrix polynomials*

Leo Taslaman, University of Manchester, UK; Yuji Nakatsukasa, University of Manchester,UK; Franc¸oise Tisseur, University of Manchester, UK

Wednesday, June 20

MS 36. **Hybrid solvers for sparse linear equations**

Organizer: Iain S Duff

STFC Rutherford Appleton Laboratory, UK

Organizer: Luc Giraud

Joint INRIA-CERFACS Laboratory, France

There is now a recognition that neither direct nor iterative methods by themselves are capable of solving some of the really large challenging sparse systems coming from large scale simulation. In this minisymposium, we consider a range of solution techniques that combine direct and iterative solution approaches and show that they are capable of solving equations with over a billion unknowns. We consider methods based on domain decomposition and block iterative methods such as the Block Cimmino. Some of the contributions also study the implementation of such approaches on parallel architectures.

11:00–11:25 *The augmented block-Cimmino distributed method*

Mohamed Zenadi, ENSEEIHT-IRIT; Iain Duff, CERFACS and RAL; Ronan Guivarch, ENSEEIHT-IRIT; Daniel Ruiz, ENSEEIHT-IRIT

11:25–11:50 *On a parallel hierarchical algebraic domain decomposition method for a large scale sparse linear solver *

Luc Giraud, Inria, joint Inria-CERFACS Lab; Emmanuel Agullo, Inria; Abdou Guermouche, Bordeaux University; Jean Roman, Inria

11:50–12:15 *A two-level Schwarz method for systems with high contrasts*

Nicole Spillane, Université Pierre et Marie Curie; Victorita Dolean, Université de Nice-Sophia Antipolis; Patrice Hauret, MFP Michelin; Fréric Nataf, Université dé Pierre et Marie Curie; Clemens Pechstein, Johannes Kepler University; Robert Scheichl, University of Bath

12:15–12:40 *A 3-level parallel hybrid preconditioner for sparse linear systems*

Erik Boman, Sandia National Labs; Siva Rajamanickam, Sandia National Labs; Michael Heroux, Sandia National Labs; Radu Popescu, EPFL, Switzerland

Wednesday, June 20

MS 37. **Optimization methods for tensor decomposition**

Organizer: Hans De Sterck

University of Waterloo

Tensor decomposition has emerged as an important technique in a variety of application domains, which include signal processing, chemometrics, datamining, neuroscience, and many more. For many applications, (approximate) tensor decomposition can naturally be posed as an optimization problem, and there is signiﬁcant current interest in how new tensor decomposition algorithms can be developed that originate from this optimization viewpoint, aiming to improve over existing approaches like alternating least-squares based methods, which are versatile but can be inefﬁcient. This mini-symposium will highlight recent results on new optimization-based algorithms for several types of tensor decomposition problems.

11:00–11:25 *Efﬁcient algorithms for tensor decompositions*

Laurent Sorber, KU Leuven; Marc Van Barel, KU Leuven; Lieven De Lathauwer, KU Leuven

11:25–11:50 *Symmetric tensor decomposition via a power method for the generalized tensor eigenproblem*

Jackson R. Mayo, Sandia National Laboratories; Tamara G. Kolda, Sandia National Laboratories

11:50–12:15 *All-at-once optimization for coupled matrix and tensor factorizations*

Evrim Acar, University of Copenhagen; Tamara G. Kolda, Sandia National Laboratories; Daniel M. Dunlavy, Sandia National Laboratories; Rasmus Bro, University of Copenhagen

12:15–12:40 *An algebraic multigrid optimization method for low-rank canonical tensor decomposition*

Killian Miller, University of Waterloo; Hans De Sterck, University of Waterloo

Wednesday, June 20

MS 38. **Generalized inverses and applications -Part I of II**

Organizer: Dragana S. Cvetkovic-Ilic

University of Nis, Serbia

Organizer: Néstor Thome

Universitat Politècnica de València, Spain

Organizer: Yimin Wei

Fudan University, P.R. China

Generalized inverses was ﬁrst introduced on operators (Fredholm (1903), Hilbert (1904)) and later on matrices (Moore (1920), Penrose (1955)). The most important fact was its conection with least-squares method. Theory, applications and computational methods of generalized inverses have been lastly developed (important monographs were written by Rao and Mitra, Ben-Israel and Greville, Campbell and Meyer, Wang, Wei and Qiao). Generalized inverses cover a wide range of mathematical areas: matrix theory, operator theory, C∗-algebras or rings. Recent studies focus on: numerical computation, reverse order law, perturbation theory, partial orders, etc. Numerous applications include areas such as: statistics, differential equations, numerical analysis, Markov chains, cryptography, control theory, coding theory, incomplete data recovery and robotics.

11:00–11:25 *The group inverse of additively modiﬁed matrices*

Nieves Castro González, Universidad Politécnica de Madrid, Spain

11:25–11:50 *The Moore-Penrose inverse of a linear combination of commuting generalized and hypergeneralized projectors*

Dragana S. Cvetkovíc-Ilíc, University of Nis, Serbia

11:50–12:15 *Generalized inverses of operators on Hilbert C∗-modules*

Dragan S. Djordjevic, University of Nis, Serbia

12:15–12:40 *Some results on the reverse order law*

Dijana Mosic, University of Nis, Serbia

Wednesday, June 20

MS 39. **Challenges for the solution and preconditioning of multiple linear systems -Part I of II**

Organizer: Eric de Sturler

Virginia Tech, USA

Organizer: Daniel B. Szyld

Temple University, USA

This minisymposium presents challenging problems and new methods for the solution and preconditioning of multiple linear systems. These include parameterized linear systems, systems with multiple shifts, slowly varying systems, and variants from a range of applications, such as acoustics, model reduction, electronic structure, time-dependent problems, and linear and nonlinear eigenvalue problems. These methods include tensor Krylov subspace methods, recycling Krylov subspaces and other subspaces, and techniques for recycling preconditioners, including AMG and other multilevel preconditioners.

11:00–11:25 *Preconditioners for sequences of shifted linear systems *

Martin B. van Gijzen, Delft University of Technology, the Netherlands; Daniel B. Szyld, Temple University, USA

11:25–11:50 *Solving sequences of linear systems with application to model reduction*

Kapil Ahuja, Max Planck Institute for Dynamics of Complex Technical Systems, Germany; Eric de Sturler, Virginia Tech, USA; Serkan Gugercin, Virginia Tech, USA; Peter Benner, Max Planck Institute for Dynamics of Complex Technical Systems, Germany

11:50–12:15 *Krylov subspace recycling for families of shifted linear systems*

Kirk M. Soodhalter, Temple University, USA; Daniel B. Szyld, Temple University, USA; Fei Xue, Temple University, USA

12:15–12:40 *Krylov subspace recycling for faster model reduction algorithms*

Peter Benner, Max Planck Institute for Dynamics of Complex Technical Systems, Germany; Lihong Feng, Max Planck Institute for Dynamics of Complex Technical Systems, Germany

Wednesday, June 20

MS 40. **Different perspectives on conditioning and numerical stability -Part I of II**

Organizer: Froilán M. Dopico

Universidad Carlos III, Spain

Organizer: Ilse C.F. Ipsen

North Carolina State University, USA

The purpose of the minisymposium is two-fold: First, to call attention to the continued importance of conditioning and numerical stability, which have been expanding far beyond the traditional direct methods for dense matrices. Second to stimulate awareness and cross fertilization of concepts and techniques among different areas. To capture the wide variety of emerging directions, the minisymposium will consist of two sessions. Different perspectives on conditioning and numerical stability will be highlighted in the following speciﬁc contexts: functions of matrices, tensors, ﬁnite elements for PDEs, fast algorithms for rank structured matrices, Krylov methods for eigenvalues, orthogonalization methods, randomized algorithms, and high relative accuracy methods.

11:00–11:25 *Highly accurate numerical linear algebra via rank revealing decompositions*

Froilán M. Dopico, Univ. Carlos III, Spain

11:25–11:50 *Stability of numerical algorithms with quasiseparable matrices*

Pavel Zhlobich, Univ. of Edinburgh, UK; Froilán M. Dopico, Univ. Carlos III, Spain; Vadim Olshevsky, Univ. of Connecticut, USA

11:50–12:15 *Gram-Schmidt orthogonalization with standard and non-standard inner product: rounding error analysis*

Miroslav Rozlozník, Academy of Sciences, Czech Rep.; Jirí Kopal, Tech. Univ. of Liberec, Czech Rep.; Alicja Smoktunowicz, Warsaw Univ. of Tech., Poland; Miroslav Tuma, Academy of Sciences, Czech Rep

12:15–12:40 *Backward stability of iterations for computing the polar decomposition*

Nicholas J. Higham, Univ. of Manchester, UK; Yuji Nakatsukasa, Univ. of Manchester, UK

Wednesday, June 20

MS 41. **Recent advances in model reduction -Part I of II**

Organizer: Athanasios C. Antoulas

Rice University, USA

Organizer: Serkan Gugercin

Virginia Polytechnic Institute, USA

Direct numerical simulation of dynamical systems has been one of very few strategies when objectives include accurate prediction or control of complex physical phenomena. However, large-scale simulations often threaten untenable demands on computational resources, and in this way provide the main motivation for model reduction: Produce a simpler reduced-order dynamical system that approximates the input-output map of the original system accurately, yet cheaply. This mini-symposium will highlight the recent advances in model reduction from data-driven model reduction techniques to optimal model reduction techniques in simulation and control referring to applications that range from ﬂuid ﬂow to circuit simulation.

11:00–11:25 *The Loewner framework in data-driven model reduction*

Athanasios C. Antoulas, Rice University

11:25–11:50 *Robust computational approaches to H2-optimal model reduction*

Christoper A. Beattie, Virginia Polytechnic Institute, USA; Serkan Gugercin, Virginia Polytechnic Institute, USA

11:50–12:15 *Reduced order modeling via frames*

Volker Mehrmann, TU Berlin; Sadegh Jokar, TU Berlin; Sarosh Quraischi, TU Berlin

12:15–12:40 *Semideﬁnite Hankel-type model reduction based on frequency response matching*

Aivar Sootla, Lund University; Anders Rantzer, Lund University; Kin Cheong Sou, KTH

Wednesday, June 20

MS 42. **Structured solution of nonlinear matrix equations and applications -Part I of II**

Organizer: Eric King-wah Chu

Monash University, Australia

Organizer: Wen-Wei Lin

National Chiao Tung University, Taiwan

Nonlinear matrix equations, such as X + BX−1A + C =0, AX2 + BX + C =0 and AX + XD − XCX + B =0, arise in many applications, such as vibration analysis, optimal control, queueing systems, nano research and economic dynamics. The exploitation of speciﬁc symmetry and structures in these equations is vital in their numerical solution, especially for large-scale problems. Much have been achieved recently, using Newton’s iteration, cyclic reduction, doubling and other methods. The minisymposium will exhibit some of these advances and achievements, as well as explore some future directions of research and applications.

11:00–11:25 *Structured solution of large-scale algebraic Riccati and nonlinear matrix equations*

Eric King-wah Chu, Monash University; Tiexiang Li, Southeast University; Wen-Wei Lin, National Chiao Tung University; Peter Chang-Yi Weng, Monash University

11:25–11:50 *Accurate solutions of nonlinear matrix equations in queueing models*

Qiang Ye, University of Kentucky, U.S.

11:50–12:15 *A numerical approach for solving nonlinear matrix equations in economic dynamics*

Matthew M. Lin, National Chung Cheng University; Moody T. Chu, North Carolina State University; Chun-Hung Kuo, North Carolina State University

12:15–12:40 *A structure-preserving doubling algorithm for quadratic eigenvalue problems arising from time-delay systems*

Tiexiang Li, Southeast University; Eric King-wah Chu, Monash University; Wen-Wei Lin, National Chiao Tung University

Wednesday, June

MS 43. **Challenges for the solution and preconditioning of multiple linear systems -Part II of II**

Organizer: Eric de Sturler

Virginia Tech, USA

Organizer: Daniel B. Szyld

Temple University, USA

This minisymposium presents challenging problems and new methods for the solution and preconditioning of multiple linear systems. These include parameterized linear systems, systems with multiple shifts, slowly varying systems, and variants from a range of applications, such as acoustics, model reduction, electronic structure, time-dependent problems, and linear and nonlinear eigenvalue problems. These methods include tensor Krylov subspace methods, recycling Krylov subspaces and other subspaces, and techniques for recycling preconditioners, including AMG and other multilevel preconditioners.

15:00–15:25 *Low-rank techniques for parameter-dependent linear systems and eigenvalue problems*

Christine Tobler, EPF Lausanne, Switzerland; Daniel Kressner, EPF Lausanne, Switzerland

15:25–15:50 *Recycling Krylov subspace information in sequences of linear systems*

Nemanja Bozovic, Bergische Universität Wuppertal

15:50–16:15 *Efﬁciently updating preconditioners in quantum Monte Carlo simulations*

Arielle Grim McNally, Virginia Tech, USA; Eric de Sturler, Virginia Tech, USA; Kapil Ahuja, Max Planck Institute for Dynamics of Complex Technical Systems, Germany; Li Ming, Virginia Tech, USA

16:15–16:40 *A domain decomposition preconditioned recycling GMRES for stochastic parabolic PDE*

Xiao-Chuan Cai, University of Colorado at Boulder, USA

Wednesday, June 20

MS 44. **Different perspectives on conditioning and numerical stability -Part II of II**

Organizer: Froilán M. Dopico

Universidad Carlos III de Madrid, Spain

Organizer: Ilse C.F. Ipsen

North Carolina State University, USA

The purpose of the minisymposium is two-fold: First, to call attention to the continued importance of conditioning and numerical stability, which have been expanding far beyond the traditional direct methods for dense matrices. Second to stimulate awareness and cross fertilization of concepts and techniques among different areas. To capture the wide variety of emerging directions, the minisymposium will consist of two sessions. Different perspectives on conditioning and numerical stability will be highlighted in the following speciﬁc contexts: functions of matrices, tensors, ﬁnite elements for PDEs, fast algorithms for rank structured matrices, Krylov methods for eigenvalues, orthogonalization methods, randomized algorithms, and high relative accuracy methods.

15:00–15:25 *Accuracy and sensitivity of Monte Carlo matrix multiplication algorithms*

John T. Holodnak, North Carolina State University, USA; Ilse C. F. Ipsen, North Carolina State University, USA

15:25–15:50 *Hyperdeterminant and the condition number of a multilinear system*

Lek-Heng Lim, University of Chicago, USA

15:50–16:15 *Condition numbers and backward errors in functional setting*

Agnieszka Miedlar, Technical University of Berlin, Germany; Mario Arioli, STFC Rutherford Appleton Laboratory, UK

16:15–16:40 *Orthogonality and stability in large-sparse-matrix iterative algorithms*

Chris Paige, McGill University, Canada; Wolfgang Wülling, W2 Actuarial & Math Services Ltd., Germany

MS 45. **Recent advances in model reduction -Part II of II**

Organizer: Athanasios C. Antoulas

Rice University, USA

Organizer: Serkan Gugercin

Virginia Polytechnic Institute and State University, USA

Direct numerical simulation of dynamical systems has been one of very few strategies when objectives include accurate prediction or control of complex physical phenomena. However, large-scale simulations often threaten untenable demands on computational resources, and in this way provide the main motivation for model reduction: Produce a simpler reduced-order dynamical system that approximates the input-output map of the original system accurately, yet cheaply. This mini-symposium will highlight the recent advances in model reduction from data-driven model reduction techniques to optimal model reduction techniques in simulation and control referring to applications that range from ﬂuid ﬂow to circuit simulation.

15:00–15:25 *Automating DEIM for nonlinear model reduction*

Danny Sorensen, Rice University

15:25–15:50 *Model reduction for optimal control problems in ﬁeld-ﬂow fractionation*

Tatjana Stykel, Universität Augsburg

15:50–16:15 *Numerical implementation of the iterative rational Krylov algorithm for optimal H2 model order reduction*

Zlatko Drmac, University of Zagreb; Christopher A. Beattie, Virginia Polytechnic Institute; Serkan Gugercin, Virginia Polytechnic Institute

16:15–16:40 *Low rank deﬂative/iterative solutions of Luré equations*

Timo Reis, Universität Hamburg; Federico Poloni, Technische Universität Berlin

Wednesday, June 20

MS 46. **Structured solution of nonlinear matrix equations and applications -Part II of II**

Organizer: Eric King-wah Chu

Monash University, Australia

Organizer: Wen-Wei Lin

National Chiao Tung University, Taiwan

Nonlinear matrix equations, such as X + BX^{-1}A + C =0, AX^{2} + BX + C =0 and AX + XD − XCX + B =0, arise in many applications, such as vibration analysis, optimal control, queueing systems, nano research and economic dynamics. The exploitation of speciﬁc symmetry and structures in these equations is vital in their numerical solution, especially for large-scale problems. Much have been achieved recently, using Newton’s iteration, cyclic reduction, doubling and other methods. The minisymposium will exhibit some of these advances and achievements, as well as explore some future directions of research and applications.

15:00–15:25 *Inertia and rank characterizations of the expressions A − BXB^{∗} − CYC*{∗} and A − BXC^{∗} ± CX^{∗}B^{∗}*

Delin Chu, National University of Singapore

15:25–15:50 *Structure-preserving Arnoldi-type algorithm for solving eigenvalue problems in leaky surface wave propagation*

Tsung-Ming Huang, National Taiwan Normal University; Wen-Wei Lin, National Chiao Tung University; Chin-Tien Wu, National Chiao Tung University

15:50–16:15 *Structure-preserving curve for symplectic pairs*

Yueh-Cheng Kuo, National University of Kaohsiung; Shih-Feng Shieh, National Taiwan Normal University

16:15–16:40 *A doubling algorithm with shift for solving a nonsymmetric algebraic Riccati equation*

Chun-Yueh Chiang, National Formosa University; Matthew M. Lin, National Chung Cheng University

Wednesday, June 20

MS 47. **Generalized inverses and applications -Part II of II**

Organizer: Dragana S. Cvetkovic-Ilic

University of Nis, Serbia

Organizer: Néstor Thome

Universitat Politècnica de València, Spain

Organizer: Yimin Wei

Fudan University, P.R. China

Generalized inverses was ﬁrst introduced on operators (Fredholm (1903), Hilbert (1904)) and later on matrices (Moore (1920), Penrose (1955)). The most important fact was its conection with least-squares method. Theory, applications and computational methods of generalized inverses have been lastly developed (important monographs were written by Rao and Mitra, Ben-Israel and Greville, Campbell and Meyer, Wang, Wei and Qiao). Generalized inverses cover a wide range of mathematical areas: matrix theory, operator theory, C∗-algebras or rings. Recent studies focus on: numerical computation, reverse order law, perturbation theory, partial orders, etc. Numerous applications include areas such as: statistics, differential equations, numerical analysis, Markov chains, cryptography, control theory, coding theory, incomplete data recovery and robotics.

15:00–15:25 *On a partial order deﬁned on certain matrices*

Néstor Thome, Universitat Politècnica de València, Spain; Araceli Hernández, Marina Lattanzi and Fabián Urquiza, Universidad Nacional de La Pampa, Argentina

15:25–15:50 *Generalized inverses and path products*

Pedro Patrício, Universidade do Minho, Portugal; R. Hartwig, N.C.S.U. Raleigh, USA

15:50–16:15 *On structured condition numbers for a linear functional of Tikhonov regularized solution*

Yimin Wei, Fudan University, P.R. China; Huaian Diao, Northeast Normal University, P.R. China

16:15–16:40 *Explicit characterization of the Drazin index*

Qingxiang Xu, Shanghai Normal University, P.R. China

Wednesday, June 20

MS 48. **Parallelization of efﬁcient algorithms**

Organizer: Matthias Bolten

University of Wuppertal, Germany

Organizer: Stefan Kunis

University of Osnabrueck, Germany

Effective discretizations and efﬁcient algorithms are necessary for modelling complex and high dimensional problems in various applications. Algorithms, scaling up to logarithmic factors linear in the problem size, typically reuse intermediate results several times and thus have a strong data dependence. We concentrate on best practice examples on modern computing architectures and aim to conclude with general guidelines for the parallelization of efﬁcient algorithms.

15:00–15:25 *A highly scalable error-controlled fast multipole method*

Ivo Kabadshow, Forschungszentrum Jülich, Supercomputing Centre

15:25–15:50 *A parallel fast Coulomb solver based on nonequispaced Fourier transforms*

Michael Pippig, Chemnitz University of Technology

15:50–16:15 *Generalized fast Fourier transforms via CUDA*

Susanne Kunis, University of Osnabrück

16:15–16:40 *Efﬁcient regularization and parallelization for sparse grid regression*

Dirk Pﬂuger, Technische Universität München

Wednesday, June 20

MS 49. **Analysis and computation on matrix manifold**

Organizer: Bruno Iannazzo

Università di Perugia, Italy

The deﬁnition and analysis of manifold structures on certain sets of matrices and tensors have proved to be a fruitful topic in pure and applied linear algebra, opening the possibility to ﬁnd the right deﬁnition of matrix geometric mean, to give new understandings and methods for classical problems as the eigencomputation or the solution of matrix equations. The power of the manifold approach has led to applications which range from elasticity to medical imaging and signal processing. Current lines of research are the theoretical analysis of matrix and tensor manifolds and the design of algorithms which exploit the manifold structures.

15:00–15:25 *Best low multilinear rank approximation of symmetric tensors by Jacobi rotations*

Pierre-Antoine Absil, Université catholique de Louvain; Mariya Ishteva, Georgia Institute of Technology; Paul Van Dooren, Université catholique de Louvain

15:25–15:50 *Differential geometry for tensors with ﬁxed hierarchical Tucker rank*

Bart Vandereycken École Polytechnique Fédérale de Lausanne; André Uschmajew, Technische Universität (TU) Berlin

15:50–16:15 *Deterministic approaches to the Karcher mean of positive deﬁnite matrices*

Yongdo Lim, Kyungpook National University

16:15–16:40 *The Karcher mean: ﬁrst and second order optimization techniques on matrix manifolds*

Ben Jeuris, Katholieke Universiteit Leuven; Raf Vandebril, Katholieke Universiteit Leuven; Bart Vandereycken, Ecole Polytechnique Fédérale de Lausanne

Wednesday, June 20

MS 50. **Advanced methods for large eigenvalue problems and their applications**

Organizer: Tetsuya Sakurai

University of Tsukuba, Japan

Organizer: Nahid Emad

University of Versailles, France

The massive increase of the number of processors into high performance computers and the number of cores into these processors make more complex the new architectures or those emerging. There is a growing need for efﬁcient numerical methods to take full advantage of the ability of these supercomputers in various computational ﬁelds. The minisymposium focuses on advanced methods for large-scale eigenvalue problems that arise in scientiﬁc and industrial areas. Emphasis are placed on the new numerical approaches to exploit the massive and multi-level parallelism of these supercomputers.

15:00–15:25 *DQDS with aggressive early deﬂation for computing singular values*

Kensuke Aishima, University of Tokyo, Japan; Yuji Nakatsukasa, University of Manchester, UK; Ichitaro Yamazaki, University of Tennessee, USA

15:25–15:50 *A scalable parallel method for large scale nonlinear eigenvalue problems*

Kazuma Yamamoto, University of Tsukuba, Japan; Tetsuya Sakurai, University of Tsukuba, Japan

15:50–16:15 *Application of the Sakurai-Sugiura method in the ﬁeld of density functional theory on highly parallel systems*

Georg Huhs, Barcelona Supercomputing Center, Spain

16:15–16:40 *MERAM for neutron physics applications using YML environment on post petascale heterogeneous architecture*

Christophe Calvin, Gif-Sur-Yvette Cedex France; Nahid Emad, University of Versailles, France; Serge Petiton, Lille University/INRIA France; Jérôme Dubois, CEA Saclay, France; Makarem Dandouna, University of Versailles, France

Thursday, June 21

MS 51. **Accurate and veriﬁed numerical computations for numerical linear algebra**

Organizer: Takeshi Ogita

Tokyo Woman’s Christian University, Japan

Organizer: Siegfried M. Rump

Hamburg University of Technology, Germany

This minisymposium is devoted to accurate and veriﬁed computations for numerical linear algebra. Such computations have become increasingly important in wide range of science and engineering, especially when requiring high reliability in solving ill-conditioned problems. Although there are many useful numerical algorithms and softwares for obtaining reliable results, they are still not widely known nor used in practical applications. The main objective of the minisymposium is to discuss several recent topics on fast, accurate and veriﬁed numerical algorithms and softwares for linear systems, eigenvalue problems, ﬂoating-point arithmetic and multiple precision arithmetic.

11:00–11:25 *Product decomposition and its applications*

Naoya Yamanaka, Waseda University; Shiníchi Oishi, Waseda University

11:25–11:50 *The MPACK: multiple precision version of BLAS and LAPACK*

Maho Nakata, RIKEN

11:50–12:15 *On eigenvalue computations of nonderogatory matrices*

Aurél Galántai, Óbuda University

12:15–12:40 *Veriﬁed solutions of sparse linear systems*

Takeshi Ogita, Tokyo Woman’s Christian University

Thursday, June 21

MS 52. **Numerical linear algebra libraries for high end computing -Part I of II**

Organizer: L. A. Drummond

Berkeley, US

Organizer: Nahid Emad

University of Versailles, France

Organizer: Jose E. Roman

Universitat Politècnica de València, Spain

The increase in computational power and the diversity of hardware have prompted the inception of linear algebra algorithms that can exploit multiple levels of concurrency to achieve larger orders of computational and problem solving scalability. Here we review the state of the art numerical linear algebra kernels for HPC. We focus on new algorithmic developments, parameterization and optimization, as well as software implementations and their deployment in emerging computer systems. Also relevant to this mini-symposium are the software reusability and library interoperability efforts to jumpstart the academic and industrial software development.

11:00–11:25 *Large-scale eigenvalue computation with PETSc and YML*

Makarem Dandouna, Univ. of Versailles, France; Nahid Emad, Univ. of Versailles, France; L.A. (Tony) Drummond, Lawrence Berkeley National Laboratory, USA

11:25–11:50 *Sparse matrix-matrix operations in PETSc*

Hong Zhang, Illinois Institute of Technology, USA; Barry Smith and PETSc Team, Argonne National Laboratory, USA

11:50–12:15 *Hierarchical QR factorization algorithms for multi-core cluster systems*

Julien Langou, Univ. of Colorado Denver, USA; Jack Dongarra and Mathieu Faverge, Univ. of Tennessee, USA; Thomas Herault, Univ. Paris-Sud, France; Yves Robert, Ecole Normale Supérieure de Lyon, France

12:15–12:40 *Towards robust numerical algorithms for exascale simulation*

Emmanuel Agullo, INRIA Bordeaux Sud-Ouest, France; Luc Giraud, Abdou Guermouche, Jean Roman and Mawussi Zounon, INRIA, France

Thursday, June 21

MS 53. **Efﬁcient preconditioners for real world applications -Part I of II**

Organizer: Massimiliano Ferronato

University of Padova, Italy

Organizer: Carlo Janna

University of Padova, Italy

Organizer: Luca Bergamaschi

University of Padova, Italy

The efﬁcient solution to sparse linear systems is quite a common issue in several real world applications and often represents the main memory-and time-consuming task in a computer simulation. In many areas of engineering and scientiﬁc computing, the solution to large sparse systems relies on iterative methods based on Krylov subspaces. Nonetheless, to become really efﬁcient Krylov solvers need appropriate preconditioning to achieve convergence in a reasonable number of iterations. Unfortunately, it is widely recognized that an optimal general-purpose preconditioner is unlikely to exist. In this minisymposium, we want to present novel scalar and parallel preconditioning techniques speciﬁcally designed for real world applications.

11:00–11:25 *A parallel factored preconditioner for non-symmetric linear systems*

Carlo Janna, Univ. of Padova, Italy; Massimiliano Ferronato and Giorgio Pini, Univ. of Padova, Italy

11:25–11:50 *Preconditioning for linear least-squares problems*

Miroslav Tuma, Academy of Sciences of the Czech Republic, Czech Republic; Rafael Bru, José Mas and José Marín, Univ. Politècnica de València, Spain

11:50–12:15 *Robust and parallel preconditioners for mechanical problems*

Kees Vuik, Technology Univ. of Delft, The Netherlands; Tom Jönsthövel and Martin van Gijzen, Technology Univ. of Delft, The Netherlands

12:15–12:40 *Block factorized forms of SPAI*

Thomas Huckle, Technische Univ. München, Germany; Matous Sedlacek, Technische Univ. München, Germany

Thursday, June 21

MS 54. **Solving ill-posed systems via signal-processing techniques-Part I of II**

Organizer: Stephen Becker

LJLL, Paris-6/CNRS, France

Organizer: Dirk Lorenz

TU Braunschweig Germany

In compressed sensing (CS) and sparse recovery problems, one seeks to solve an underdetermined or ill-posed equation by exploiting prior knowledge such as sparsity. These signal-processing problems have beneﬁted from established results on optimization and greedy methods, but they have also introduced many novel results which are widely-applicable even outside the target applications. This minisymposium focuses on the computational, algorithmical and numerical aspects of these techniques. We bring together researchers who work on a wide variety of approaches, e.g. subgradient methods, semismooth methods or proximal methods.

11:00–11:25 *Sequential updates for L1 minimization: sparse Kalman ﬁltering, reweighted L1, and more*

Justin Romberg, Georgia Tech; M. Salman Asif, Georgia Tech.

11:25–11:50 *Solving basis pursuit: infeasible-point subgradient algorithm, computational comparison, and improvements*

Andreas Tillmann, Technische Universität Braunschweig; Dirk Lorenz, Technische Universität Braunschweig; Marc Pfetsch, Technische Universität Braunschweig

11:50–12:15 *Semismooth Newton methods with multi-dimensional ﬁlter globalization for l_{1} optimization*

Andre Milzarek, Technische Universität Muenchen; Michael Ulbrich, Technische Universität Muenchen

12:15–12:40 *Improved ﬁrst-order methods: how to handle constraints, non-smoothness, and slow convergence*

Stephen Becker, Paris-6/CNRS; Jalal Fadili, CNRS-ENSICAEN-Université de Caen; Emmanuel Candès, Stanford University; Michael Grant, CVX Research

Thursday, June 21

MS 55. **Max-algebra -Part I of II**

Organizer: Hans Schneider

Univ. of Wisconsin, USA

Organizer: Peter Butkovic

Univ. of Birmingham, UK

Max-algebra (also known as tropical algebra) and its generalization to idempotent algebra are rapidly evolving areas of algebra, designed to solve problems which are non-linear in classical algebra but become linear in max-algebra. Such problems may be found in mathematics, operational research, computer science and engineering. While 25 years ago there were only isolated researchers in this area, since 1995 we have seen remarkable expansion following a number of advances and applications in areas as diverse as algebraic geometry, phylogenetics and railway scheduling. This is the ﬁrst of the two mini-symposia that provide eight state-of-the-art research presentations.

11:00–11:25 *Tropical bounds for the eigenvalues of structured matrices*

Marianne Akian, INRIA Saclay-IIle-de-France; Stephane Gaubert, INRIA Saclay–IIle-de-France; Meisam Sharify, INRIA Saclay–IIle-de-France

11:25–11:50 *Sensitivity in extremal systems of linear equations and inequalities*

Karel Zimmermann, Charles University Prague; Martin Gavalec, University of Hradec Králové

11:50–12:15 *Multiplicative structure of tropical matrices*

Mark Kambites, University of Manchester

12:15–12:40 *Transience bounds for matrix powers in max algebra*

Sergeı Sergeev, University of Birmingham

Thursday, June 21

MS 56. **Eigenvalue perturbations and pseudospectra -Part I of II**

Organizer: Daniel Kressner

EPF Lausanne, Switzlerland

Organizer: Julio Moro

Universidad Carlos III de Madrid, Spain

The behavior of matrix eigenvalues under perturbations is a classical topic in (numerical) linear algebra and has been under continuous investigation for several decades. The aim of this minisymposium is to highlight some recent exciting developments in this area. For example, in applications from systems and control theory it is often more realistic to impose a certain structure on the perturbations. Signiﬁcant progress has been made in analyzing and computing the corresponding structured eigenvalue condition numbers and pseudospectra. For both, structured and unstructured pseudospectra, a more reﬁned understanding of analytical properties has led to improved algorithms for computing pseudospectral quantities, such as the pseudospectral abscissa or the H∞ norm. Finally, the localization of Ritz value is an impressive recent example for the use of pseudospectra in understanding the convergence of Krylov subspace methods.

11:00–11:25 *Inclusion theorems for pseudospectra of block triangular matrices*

Michael Karow, TU Berlin, Germany

11:25–11:50 *Conjectures on pseudospectra of matrices*

Juan-Miguel Gracia, The Univ. of the Basque Country, Spain

11:50–12:15 *Sensitivity of eigenvalues of an unsymmetric tridiagonal matrix*

Carla Ferreira, Univ. of Minho, Portugal; Beresford Parlett, Univ. of California, USA; Froilán Dopico, Univ. Carlos III de Madrid, Spain

12:15–12:40 *First order structured perturbation theory for eigenvalues of skew-adjoint matrices*

Julio Moro, Univ. Carlos III de Madrid, Spain; María J. Peláez, Univ. Cat. del Norte, Chile

Thursday, June 21

MS 57. **Numerical linear algebra and optimization in imaging applications -Part I of II**

Organizer: Julianne Chung

University of Texas at Arlington, USA

Organizer: Roummel Marcia

University of California, USA

Image processing is widely used in many of today’s practical applications, from medicine and astronomy to ecology and security. Oftentimes, the underlying problem in these applications is an inverse problem, posing many challenges such as scale and ill-posedness. While many numerical methods have been developed to address these challenges, recent advances in imaging techniques require faster and more robust methods. Numerical linear algebra and numerical optimization continue to play a vital role in the development of these algorithms, and this two-session minisymposium will highlight the latest contributions. A diversity of problems, algorithms, and applications will be addressed.

11:00–11:25 *Some numerical linear algebra and optimization problems in spectral imaging*

Robert Plemmons, Wake Forest University; J. Erway, Wake Forest University; N. Gillis, University of Waterloo; X. Hu, Wake Forest University; M. Ng, Hong Kong Baptist University; P. Pauca, Wake Forest University; S. Prasad, University of New Mexico; J. Zhang, Wake Forest University; Q. Zhang, Wake Forest University

11:25–11:50 *Image restoration via constrained optimization: an approach using feasible direction methods*

Germana Landi, University of Bologna

11:50–12:15 *On the solution of linear systems in Newton-type methods for image reconstruction*

Elena Piccolomini, University of Bologna

12:15–12:40 *Alternating direction optimization for convex inverse problems with application to imaging and hyperspectral unmixing*

Jose Bioucas-Dias, Technical University of Lisbon

Thursday, June 21

MS 58. **Parametric eigenvalue problems -Part I of II**

Organizer: K. Meerbergen

K.U. Leuven, Belgium

Organizer: W. Michiels

K.U. Leuven, Belgium

Organizer: C. Lecomte

University of Southampton, University Road, Highﬁeld, UK

In applications, matrices, and thus also their eigenvalues, depend on physical parameters. The eigenvalues are usually continuous functions of the parameters, a property, which is used by, e.g., numerical methods for determining pseudospectra, the distance to instability, Hopf bifurcations, etc. In this minisymposium, we bring together a collection of talks with very different viewpoints, methods and applications, related to parametric eigenvalue problems.

11:00–11:25 *Computing double eigenvalues via the two-parameter eigenvalue problem*

Bor Plestenjak, University of Ljubljana, Slovenia

11:25–11:50 *Lyapunov inverse iteration for identifying Hopf bifurcations in models of incompressible ﬂow*

Alastair Spence, University of Bath, UK; Howard Elman, University of Maryland, USA; Karl Meerbergen, K.U.Leuven, Belgium; Minghao Wu, University of Maryland, USA

11:50–12:15 *A quadratically convergent algorithm for matrix distance problems*

Melina Freitag, University of Bath, UK; Alastair Spence, University of Bath, UK

12:15–12:40 *A real Jacobi-Davidson algorithm for the 2-real-parameter eigenvalue problem*

Christian Schröder, TU Berlin, Germany

Thursday, June 21

MS 59. **Structured matrix computations -Part I of II**

Organizer: Jianlin Xia

Purdue University, US

Organizer: Xiaoye S. Li

Lawrence Berkeley National Laboratory, USA

Structured matrices have been widely used to solve large matrix problems, PDEs, integral equations, etc. In recent developments, rank structured methods with high efﬁciency, stability, and scalability are proposed, and are suitable for large-scale scientiﬁc computing. In this minisympisium, a variety of matrix structures are discussed, including hierarchical, semiseparable, and quasiseparable matrices. Their applications to large control, imaging, geophysics, and engineering problems are illustrated. New research directions for structured matrices are discussed, such as innovative structured factorization, preconditioning, and eigensolution techniques.

11:00–11:25 *Factorization of H2-matrices*

Steffen Börm, Institut für Informatik, Christian-Albrechts-Universität zu Kiel

11:25–11:50 *The polynomial root ﬁnding problems and quasiseparable representations of unitary matrices*

Yuli Eidelman, Tel-Aviv University, Israel

11:50–12:15 *A fast direct solver for structured matrices arising from non-oscillatory integral equations*

Kenneth L. Ho, New York University, USA; Leslie Greengard, New York University

12:15–12:40 *Multivariate orthogonal polynomials and inverse eigenvalue problems*

Matthias Humet, University of Leuven; Marc Van Barel, University of Leuven

Thursday, June 21

MS 60. **Numerical linear algebra libraries for high end computing -Part II of II**

Organizer: L. A. Drummond

Berkeley, US

Organizer: Nahid Emad

University of Versailles, France

Organizer: Jose E. Roman

Universitat Politècnica de València, Spain

The increase in computational power and the diversity of hardware have prompted the inception of linear algebra algorithms that can exploit multiple levels of concurrency to achieve larger orders of computational and problem solving scalability. Here we review the state of the art numerical linear algebra kernels for HPC. We focus on new algorithmic developments, parameterization and optimization, as well as software implementations and their deployment in emerging computer systems. Also relevant to this mini-symposium are the software reusability and library interoperability efforts to jumpstart the academic and industrial software development.

15:00–15:25 *Thick-restart Lanczos methods for symmetric-indeﬁnite generalized eigenproblems in SLEPc*

Jose E. Roman, Universitat Politècnica de València, Spain; Carmen Campos, Universitat Politècnica de València, Spain

15:25–15:50 *Parametric approach to smart-tuning and auto-tuning of the DOE ACTS collection*

L. A. (Tony) Drummond, Lawrence Berkeley National Laboratory, USA; O A. Marques, Lawrence Berkeley National Laboratory, USA

15:50–16:15 *Trilinos: foundational libraries that enable next-generation computing*

Heidi Thornquist, Sandia National Laboratory, USA; The Trilinos Development Team, Sandia National Laboratory, USA

16:15–16:40 *Rethinking distributed dense Linear Algebra*

Jack Poulson, The University of Texas at Austin, USA

Thursday, June 21

MS 61. **Efﬁcient preconditioners for real world applications -Part II of II**

Organizer: Massimiliano Ferronato

University of Padova, Italy

Organizer: Carlo Janna

University of Padova, Italy

Organizer: Luca Bergamaschi

University of Padova, Italy

The efﬁcient solution to sparse linear systems is quite a common issue in several real world applications and often represents the main memory-and time-consuming task in a computer simulation. In many areas of engineering and scientiﬁc computing, the solution to large sparse systems relies on iterative methods based on Krylov subspaces. Nonetheless, to become really efﬁcient Krylov solvers need appropriate preconditioning to achieve convergence in a reasonable number of iterations. Unfortunately, it is widely recognized that an optimal general-purpose preconditioner is unlikely to exist. In this minisymposium, we want to present novel scalar and parallel preconditioning techniques speciﬁcally designed for real world applications.

15:00–15:25 *Relaxed mixed constraint preconditioners for ill-conditioned symmetric saddle point linear systems*

Luca Bergamaschi, Univ. of Padova, Italy; Ángeles Martínez, Univ. of Padova, Italy

15:25–15:50 *Chebychev acceleration of iterative reﬁnement*

Jennifer Scott, Science and Technology Faculties Council, UK; Mario Arioli, Science and Technology Faculties Council, UK

15:50–16:15 *Parallel deﬂated GMRES with the Newton basis*

Désiré Nuentsa Wakam, INRIA, Bordeaux, France; Jocelyne Erhel, INRIA, Rennes, France

16:15–16:40 *Rank-k updates of incomplete Sherman-Morrison preconditioners*

José Marín, Universidad Politécnica de Valencia, Spain; Juana Cerdán and José Mas, Universitat Politècnica de València, Spain

Thursday, June 21

MS 62. **Solving ill-posed systems via signal-processing techniques-Part II of II**

Organizer: Stephen Becker

LJLL, Paris-6/CNRS, France

Organizer: Dirk Lorenz

TU Braunschweig, Germany

In compressed sensing (CS) and sparse recovery problems, one seeks to solve an underdetermined or ill-posed equation by exploiting prior knowledge such as sparsity. These signal-processing problems have beneﬁted from established results on optimization and greedy methods, but they have also introduced many novel results which are widely-applicable even outside the target applications. This minisymposium focuses on the computational, algorithmical and numerical aspects of these techniques. We bring together researchers who work on a wide variety of approaches, e.g. subgradient methods, semismooth methods or proximal methods.

15:00–15:25 *Effects of prox parameter selection strategies in exact and inexact ﬁrst-order methods for compressed sensing and other composite optimization problems*

Katya Scheinberg, Lehigh University; Donald Goldfarb, Columbia University; Shiqian Ma, University of Minnesota

15:25–15:50 *An adaptive inverse scale space method for compressed sensing*

Martin Benning, University of Münster; Michael Möllerr, University of Münster; Pia Heins, University of Münster

15:50–16:15 *CGSO for convex problems with polyhedral constraints*

Sahar Karimi, University of Waterloo; Stephen Vavasis, University of Waterloo

16:15–16:40 *Phase-retrieval using explicit low-rank matrix factorization*

Ewout van den Berg, Stanford University; Emmanuel Candès, Stanford University

Thursday, June 21

MS 63. **Max-algebra -Part II of II**

Organizer: Hans Schneider

Univ. of Wisconsin, USA

Organizer: Peter Butkovic

Univ. of Birmingham, UK

Max-algebra (also known as tropical algebra) and its generalization to idempotent algebra are rapidly evolving areas of algebra, designed to solve problems which are non-linear in classical algebra but become linear in max-algebra. Such problems may be found in mathematics, operational research, computer science and engineering. While 25 years ago there were only isolated researchers in this area, since 1995 we have seen remarkable expansion following a number of advances and applications in areas as diverse as algebraic geometry, phylogenetics and railway scheduling. This is the second of the two mini-symposia that provide eight state-of-the-art research presentations.

15:00–15:25 *Three-dimensional convex polyhedra tropically spanned by four points*

María Jesús de la Puente, Universidad Complutense de Madrid, Spain; Adrián Jiménez, Universidad Complutense de Madrid, Spain

15:25–15:50 *Algorithmic problems in tropical convexity*

Xavier Allamigeon, INRIA Saclay–IIle-de-France and CMAP, Ecole Polytechnique; Stéphane Gaubert, INRIA Saclay–Ile-de-France and CMAP Ecole Polytechnique; Eric Goubault, CEA Saclay Nano-INNOV; Ricardo D. Katz, Universidad Nacional de Rosario

15:50–16:15 *On the weak robustness of interval fuzzy matrices*

Ján Plávka, Technical University Kosice; Martin Gavalec, University of Hradec Králové

16:15–16:40 *Weakly stable matrices *

Peter Butkovic, University of Birmingham

Thursday, June 21

MS 64. **Eigenvalue perturbations and pseudospectra -Part II of II**

Organizer: Daniel Kressner

EPF Lausanne, Switzlerland

Organizer: Julio Moro

Universidad Carlos III de Madrid, Spain

The behavior of matrix eigenvalues under perturbations is a classical topic in (numerical) linear algebra and has been under continuous investigation for several decades. The aim of this minisymposium is to highlight some recent exciting developments in this area. For example, in applications from systems and control theory it is often more realistic to impose a certain structure on the perturbations. Signiﬁcant progress has been made in analyzing and computing the corresponding structured eigenvalue condition numbers and pseudospectra. For both, structured and unstructured pseudospectra, a more reﬁned understanding of analytical properties has led to improved algorithms for computing pseudospectral quantities, such as the pseudospectral abscissa or the H∞ norm. Finally, the localization of Ritz value is an impressive recent example for the use of pseudospectra in understanding the convergence of Krylov subspace methods.

15:00–15:25 *Ritz value localization for non-Hermitian matrices*

Mark Embree, Rice University, USA; Russell Carden, Rice University

15:25–15:50 *Optimization of eigenvalues of Hermitian matrix functions*

Emre Mengi, Koç University, Turkey; Mustafa Kiliç, Koç University, Turkey; E. Alper Yıldırim, Koç University, Turkey

15:50–16:15 *Algorithms for approximating the H∞ norm*

Michael Overton, New York University, USA; Mert Gürbüzbalaban, New York University, USA; Nicola Guglielmi, University of L’Aquila, Italy

16:15–16:40 *Reduced basis methods for computing pseudospectral quantities*

Daniel Kressner, EPF Lausanne, Switzerland

Thursday, June 21

MS 65. **Numerical linear algebra and optimization in imaging applications -Part II of II**

Organizer: Julianne Chung

University of Texas at Arlington, USA

Organizer: Roummel Marcia

University of California, USA

Image processing is widely used in many of today’s practical applications, from medicine and astronomy to ecology and security. Oftentimes, the underlying problem in these applications is an inverse problem, posing many challenges such as scale and ill-posedness. While many numerical methods have been developed to address these challenges, recent advances in imaging techniques require faster and more robust methods. Numerical linear algebra and numerical optimization continue to play a vital role in the development of these algorithms, and this two-session minisymposium will highlight the latest contributions. A diversity of problems, algorithms, and applications will be addressed.

15:00–15:25 *A recursion relation for solving L-BFGS systems with diagonal updates*

Jennifer Erway, Wake Forest University; R. Marcia, University of California, USA

15:25–15:50 *Wavefront gradients reconstruction using l^{1} − l^{p} models*

Raymond Chan, The Chinese University of Hong Kong; X. Yuan, Hong Kong Baptist University; W. Zhang, Nanjing University

15:50–16:15 *Edge-preserving image enhancement via blind deconvolution and upsampling operators*

Antonio Marquina, University of Valencia, Spain; S. Osher, UCLA; S. Joshi, UCLA

16:15–16:40 *A new hybrid-optimization method for large-scale, non-negative, full regularization*

Marielba Rojas, Delft University of Technology, The Netherlands; J. Guerrero, Carabobo University, Venezuela; M. Raydan, Simón Bolívar University, Venezuela

Thursday, June 21

MS 66.** Parametric eigenvalue problems -Part II of II**

Organizer: K. Meerbergen

K.U. Leuven, Belgium

Organizer: W. Michiels

K.U. Leuven, Belgium

Organizer: C. Lecomte

University of Southampton, University Road, Highﬁeld, UK

In applications, matrices, and thus also their eigenvalues, depend on physical parameters. The eigenvalues are usually continuous functions of the parameters, a property, which is used by, e.g., numerical methods for determining pseudospectra, the distance to instability, Hopf bifurcations, etc. In this minisymposium, we bring together a collection of talks with very different viewpoints, methods and applications, related to parametric eigenvalue problems.

15:00–15:25 *A subspace optimization technique for the generalized minimal maximal eigenvalue problem*

Jeroen De Vlieger, K.U.Leuven, Belgium; Karl Meerbergen, K.U.Leuven, Belgium

15:25–15:50 *An iterative method for computing pseudospectral abscissa and stability radii for nonlinear eigenvalue problems*

Wim Michiels, K.U.Leuven, Belgium; Nicola Guglielmi, Università dell’Aquila, Italy

15:50–16:15 *Statistical pseudospectrum and eigenvalue robustness to rank-one disturbance*

Christophe Lecomte, University of Southampton, UK; Maryam Ghandchi Tehrani, University of Southampton, UK

Thursday, June 21

MS 67. **Structured matrix computations -Part II of II**

Organizer: Jianlin Xia

Purdue University, US

Organizer: Xiaoye S. Li

Lawrence Berkeley National Laboratory, USA

Structured matrices have been widely used to solve large matrix problems, PDEs, integral equations, etc. In recent developments, rank structured methods with high efﬁciency, stability, and scalability are proposed, and are suitable for large-scale scientiﬁc computing. In this minisympisium, a variety of matrix structures are discussed, including hierarchical, semiseparable, and quasiseparable matrices. Their applications to large control, imaging, geophysics, and engineering problems are illustrated. New research directions for structured matrices are discussed, such as innovative structured factorization, preconditioning, and eigensolution techniques.

15:00–15:25 *Randomized numerical matrix computations with applications*

Victor Pan, University of New York, USA

15:25–15:50 *Massively parallel structured direct solver for the equations describing time-harmonic seismic waves*

Maarten V. de Hoop, Purdue University; Shen Wang, Purdue University; Jianlin Xia, Purdue University; Xiaoye S. Li, Lawrence Berkeley National Laboratory

15:50–16:15 *Accelerating electronic structure calculation with pole expansion plus selected inversion method*

Lin Lin and Chao Yang, Lawrence Berkeley National Laboratory

16:15–16:40 *Randomized direct solvers *

Jianlin Xia, Purdue University; Maarten V. de Hoop, Purdue University; Xiaoye S. Li, Lawrence Berkeley National Laboratory; Shen Wang, Purdue University

Friday, June 22

MS 68. **Linear algebra for structured eigenvalue computations arising from (matrix) polynomials**

Organizer: L. Gemignani

Univ. of Pisa, Italy

Organizer: R. Vandebril

KU Leuven, Belgium

Structured eigenvalue problems emerge in applications in science and engineering. Structure is induced by the original problem or introduced by linearization. The aim is to develop theory (canonical forms, perturbation theory) and numerical methods (linearizations and iterations) respecting structure and spectral properties, aiming at faster and more accurate solutions. A (generalized) companion matrix results from linearizing a (particular) polynomial root ﬁnding problem into a matrix eigenvalue problem. In this mini we will unite researchers interested in structured eigenvalue problems arising from (matrix) polynomials with a view to surveying the state-of-the-art and reporting on recent advances and challenges.

11:00–11:25 *A QR algorithm with generator compression for structured eigenvalue computation*

P. Boito, University of Limoges; Y. Eidelman, Tel-Aviv Univ.; L. Gemignani, Univ. of Pisa; I. Gohberg, Tel-Aviv Univ.

11:25–11:50 *Quadratic realizability for structured matrix polynomials*

D.S. Mackey, Western Michigan University; F. De Teáan and F.M. Dopico, Univ. Carlos III de Madrid; F. Tisseur, The University of Manchester

11:50–12:15 *Fast computation of zeros of a polynomial*

D.S. Watkins, Washington State Univ.; J.L. Aurentz, Washington State Univ.; R. Vandebril, KU Leuven

12:15–12:40 *Eigenvector recovery of linearizations and the condition number of eigenvalues of matrix polynomials*

F. De Terán, Univ. Carlos III de Madrid; M.I. Bueno, Univ. of California at Santa Barbara; F.M. Dopico, Univ. Carlos III de Madrid; D.S. Mackey, Western Michigan Univ.

Friday, June 22

MS 69. **Advances in sparse matrix factorization**

Organizer: A. Buttari

CNRS-IRIT, Toulouse, France

Organizer: S. Toledo

Tel-Aviv University, Israel

The ever increasing size of scientiﬁc problems and the fast pace at which computing architectures evolve force researchers to continuously update or rethink their algorithms and programming models. Sparse, direct solvers are no exception in this trend. This minisymposium will present recent advances, as well as the new and future challenges involving sparse, direct solvers. Speciﬁcally, it will focus on novel numerical methods for revealing mathematical properties of sparse matrices, algorithms for reducing the complexity and improving the efﬁciency of direct solvers as well as programming models for exploiting the computational power of modern, heterogeneous computers.

11:00–11:25 *A Sparse inertia-revealing factorization*

Alex Druinsky, Tel-Aviv University; S. Toledo, Tel-Aviv University

11:25–11:50 *Multifrontal factorization on heterogeneous multicore systems*

Bob Lucas, Univ. of Southern California; Roger Grimes, Livermore Software Technology Corp.; John Tran, Univ. of Southern California; Gene Wagenbreth, Univ. of Southern California

11:50–12:15 *Towards an optimal parallel approximate sparse factorization algorithm using hierarchically semi-separable structures*

Xiaoye S. Li, Lawrence Berkeley National Laboratory; Shen Wang, Purdue Univ.; Jianlin Xia, Purdue Univ.; Maarten V. de Hoop, Purdue Univ.

12:15–12:40 *Improving multifrontal methods by means of low-rank approximation techniques*

Clement Weisbecker, Univ. of Toulouse; Patrick Amestoy, Univ. of Toulouse; Cleve Ashcraft, LSTC, Livermore; Olivier Boiteau, EDF R&D, Clamart; Alfredo Buttari, CNRS-IRIT; Jean-Yves L’Excellent, INRIA-LIP(ENS Lyon)

Friday, June 22

MS 70. **Accurate algorithms and applications**

Organizer: Roberto Barrio

Universidad de Zaragoza, Spain

Organizer: Siegfried M. Rump

Hamburg University of Technology, Germany

At the present time, IEEE 64-bit ﬂoating-point arithmetic is sufﬁciently accurate for most scientiﬁc applications. However, for a rapidly growing body of important scientiﬁc computing applications, a higher level of numeric precision is required. Such calculations are facilitated by high-precision software packages. Another situation that requires new algorithms is the numerical solution in double precision of ill-posed problems in scientiﬁc applications. One alternative, related with some high-precision algorithms are the so called compensated algorithms. This minisymposium presents a survey of recent applications of these techniques and provides some analysis of their numerical requirements.

11:00–11:25 *High precision and accurate algorithms in Physics and Mathematics*

Roberto Barrio Universidad de Zaragoza, Spain; David H. Bailey, Lawrence Berkeley National Laboratory, USA; Jonathan M. Borwein, Univ. of Newcastle, Australia; Sergio Serrano, Univ. de Zaragoza, Spain

11:25–11:50 *Accurate evaluation of 1D and 2D polynomials in Bernstein form*

Hao Jiang, Universite of Pierre et Marie Curie, Paris, France; Roberto Barrio, Univ. de Zaragoza, Spain

11:50–12:15 *Some issues related to double roundings*

Jean-Michel Muller, CNRS, Université de Lyon, France; Erik Martin-Dorel, Université de Lyon, France; Guillaume Melquiond, INRIA, Université Paris Sud, France

12:15–12:40 *Error bounds for ﬂoating-point summation and dot product*

Siegfried M. Rump, Hamburg University of Technology, Germany and Waseda University, Japan

Friday, June 22

MS 71. **Theoretical and applied aspects of graph Laplacians**

Organizer: Shaun Fallat

University of Regina, CANADA

Organizer: Steve Kirkland

Hamilton Institute, National University of Ireland Maynooth, Ireland

The Laplacian matrix associated with a graph has its roots in the analysis of networks (beginning with Kirchoff) and has become a pillar for research in combinatorial matrix theory, informing new applications in both combinatorics and numerical linear algebra. Speciﬁcally, there has been a fruitful investigation (beginning with Fielder in the 1970s) of the relationship between the topological structure of a graph and the spectral properties of the corresponding Laplacian matrix. This minisymposium will pursue that theme by exploring, in theoretical and applied settings, connections between the algebraic properties of Laplacian matrices and the nature of the underlying networks.

11:00–11:25 *Potential theory for perturbed Laplacian of ﬁnite networks*

Margarida Mitjana, Universitat Politécnica de Catalunya; E. Bendito, A. Carmona and A.M. Encinas, Universitat Politécnica de Catalunya

11:25–11:50 *Subclasses of graphs with partial ordering with respect to the spectral radius of generalized graph Laplacians.*

Josef Leydold, WU Vienna University of Economics and Business; Türker Bıyıkoglu, Isık University

11:50–12:15 *Some new results on the signless Laplacian of graphs*

Slobodan K. Simic, Mathematical Institute of Serbian Academy of Science and Arts

12:15–12:40 *Graph bisection from the principal normalized Laplacian eigenvector*

Dragan Stevanovic, University of Primorska, UP IAM and University of Nis. PMF

Friday, June 22

MS 72. **Linear techniques for solving nonlinear equations**

Organizer: Vicente F. Candela Pomares and Rosa M. Peris Sancho

University of Valencia, Spain

A great deal of nonlinear problems are solved by means of linearization processes, where most of the difﬁculties of nonlinearity are smoothed but inherited by the linear models. Thus, from deconvolution (in image processing), to optimization or ill-posed equations, linear models are developed in order to simplify the problems without losing quality. In this minisymposium we present some of the recent techniques devised to get best performance of linear schemes applied to nonlinear fractional deconvolution, nonlinear variational problems, ill-posedness or nonlinear equations in general, relating both linear and nonlinear worlds.

11:00–11:25 *A Gauss-Seidel process in iterative methods for solving nonlinear equations*

José Gutiérrez, Universidad de la Rioja; A. Magreñán, University of La Rioja; J. L. Varona, University of La Rioja

11:25–11:50 *A greedy algorithm for convergence of a fractional blind deconvolution*

Vicente F. Candela, Universidad de Valencia; Pantaleón David Romero Sánchez, Universidad CEU-Cardenal Herrera

11:50–12:15 *Overview of iterative methods using a variational approch*

Sonia Busquier Sáez, Universidad Politécnica de Cartagena; S. Amat, U.P. Cartagena; P. Pedregal, Universidad de Castilla La Mancha

12:15–12:40 *Iterative methods for ill-conditioned problems*

Rosa M. Peris Sancho, University of Valencia; Vicente F. Candela, University of Valencia

Friday, June 22

MS 73. **Algebraic Riccati equations associated with M-matrices: numerical solution and applications**

Organizer: Beatrice Meini

University of Pisa, Italy

Algebraic Riccati equations are a class of matrix equations which model different real world problems. The interest in Riccati equations associated with M-matrices is recent, and is motivated by the relevant applications in ﬂuid queues, stochastic processes and transport theory. The aim of this minisymposium is to bring together people working in applications and people working in numerical linear algebra. In fact, researchers interested in applications are a source of challenging problems; people with expertise in numerical linear algebra can provide highly efﬁcient solution methods. The minisymposium should bring synergetic beneﬁts both to the theoretical and the applied scientiﬁc community.

11:00–11:25 *Monotone convergence of Newton-like methods for M-matrix algebraic Riccati equations*

Chun-Hua Guo, University of Regina, Canada

11:25–11:50 *Accurate solution of M-matrix algebraic Riccati equation by ADDA: alternating-directional doubling algorithm*

Ren-Cang Li, University of Texas at Arlington, USA

11:50–12:15 *When ﬂuid becomes Brownian: the morphing of Riccati into quadratic equations*

Giang Nguyen, Université Libre de Bruxelles, Belgium

12:15–12:40 *Analyzing multi-type queues with general customer impatience using Riccati equations*

Benny Van Houdt, University of Antwerp, Antwerpen, Belgium

Friday, June 22

MS 74. **Recent advances in the numerical solution of large scale matrix equations**

Organizer: Valeria Simoncini

Università di Bologna, Italy

Organizer: Daniel B. Szyld

Temple University, Philadelphia, USA

Linear and quadratic matrix equations arise in many areas in science and engineering. Very often these equations stem from the discretization of problems involving, possibly three-dimensional, partial differential equations, and thus the matrices have very large dimensions. Therefore, the determination of low rank approximations to the sought-after matrix solution is mandatory. In this minisymposium we present new advances on numerical techniques that speciﬁcally address the solution to popular linear and quadratic equations, such as Lyapunov-type and Riccati-type algebraic equations.

11:00–11:25 *Hierarchical and Multigrid methods for matrix and tensor equations*

Lars Grasedyck, Inst. for Geometry and Practical Mathematics, RWTH Aachen

11:25–11:50 *A Survey on Newton-ADI based solvers for large scale AREs*

Jens Saak, Max Planck Institute for Dynamics of Complex Technical Systems and Chemnitz University of Technology, DE

11:50–12:15 *An invariant subspace method for large-scale algebraic Riccati and Bernoulli equations*

Luca Amodei, Université Paul Sabatier, France; Jean-Marie Buchot, Université Paul Sabatier, France

12:15–12:40 *Delay Lyapunov equations and model order reduction of time delay systems*

Tobias Damm, University of Bayreuth; Elias Jarlebring KTH, Stockholm; Wim Michiels K.U. Leuven, BE

Friday, June 22

MS 75. **Points that minimize potential functions**

Organizer: Martin Ehler

Helmholtz Zentrum Muenchen, Germany

Organizer: Johann S. Brauchart, Postdoctoral Fellow

The University of New South Wales, Australia

Point distributions that minimize or maximize speciﬁc potential functions are used in numerical integration, coding theory, image dithering, and statistical design. Our goal is to investigate optimal conﬁgurations on the unit sphere and other manifolds for a range of such functions, including determinants as well as frame-and Riesz-potentials. Minimizers of the frame-potential are analytically characterized by approximation properties of linearly dependent sets. Extrema of determinants and Riesz-potentials on special manifolds are found numerically, and similar approaches yield quasi-uniform samplings on compact manifolds. We aim to identify common themes between research ﬁelds, which use extremal conﬁgurations, to facilitate synergistic effects.

11:00–11:25 *Discretizing compact manifolds with minimal energy*

Johann S. Brauchart, The University of New South Wales

11:25–11:50 *Well conditioned spherical designs and potential functions*

Robert S. Womersley, The University of New South Wales

11:50–12:15 *Probabilistic frames in the 2-Wasserstein metric*

Kasso Okoudjou, The University of Maryland

12:15–12:40 *Numerical minimization of potential energies on speciﬁc manifolds*

Manuel Gräf, The Chemnitz University of Technology