Text only print
Full screen print
Numerical Analysis Seminar
Seminar information archive ～12/30｜Next seminar｜Future seminars 12/31～
| Date, time & place | Tuesday 16:30 - 18:00 002Room #002 (Graduate School of Math. Sci. Bldg.) |
|---|---|
| Organizer(s) | Norikazu Saito, Takahito Kashiwabara |
Seminar information archive
2018/10/15
16:50-18:20 Room #002 (Graduate School of Math. Sci. Bldg.)
Takeyuki Nagasawa (Saitama University)
Möbius invariant discretizations and decomposition of the Möbius energy (Japanese)
Takeyuki Nagasawa (Saitama University)
Möbius invariant discretizations and decomposition of the Möbius energy (Japanese)
2018/07/31
14:00-15:00 Room #056 (Graduate School of Math. Sci. Bldg.)
Jichun Li (University of Nevada Las Vegas)
Recent advances on numerical analysis and simulation of invisibility cloaks with metamaterials (English)
Jichun Li (University of Nevada Las Vegas)
Recent advances on numerical analysis and simulation of invisibility cloaks with metamaterials (English)
[ Abstract ]
In the June 23, 2006's issue of Science magazine, Pendry et al. and Leonhardt independently published their seminar papers on electromagnetic cloaking. Since then, there is a growing interest in using metamaterials to design invisibility cloaks. In this talk, I will first give a brief introduction to invisibility cloaks with metamaterials, then I will focus on some time-domain cloaking models we studied in the last few years. Well-posedness study and time-domain finite element method for these models will be presented. I will conclude the talk with some open issues.
In the June 23, 2006's issue of Science magazine, Pendry et al. and Leonhardt independently published their seminar papers on electromagnetic cloaking. Since then, there is a growing interest in using metamaterials to design invisibility cloaks. In this talk, I will first give a brief introduction to invisibility cloaks with metamaterials, then I will focus on some time-domain cloaking models we studied in the last few years. Well-posedness study and time-domain finite element method for these models will be presented. I will conclude the talk with some open issues.
2018/07/26
16:00-17:30 Room #056 (Graduate School of Math. Sci. Bldg.)
Takahito Kashiwabara (University of Tokyo)
$L^\infty$ error estimates of the finite element method for elliptic and parabolic equations with the Neumann boundary condition in smooth domains (日本語)
Takahito Kashiwabara (University of Tokyo)
$L^\infty$ error estimates of the finite element method for elliptic and parabolic equations with the Neumann boundary condition in smooth domains (日本語)
2018/07/10
16:50-18:20 Room #126 (Graduate School of Math. Sci. Bldg.)
Junichi Matsumoto (AIST)
Free surface flow using orthogonal basis bubble function finite element method (Japanese)
Junichi Matsumoto (AIST)
Free surface flow using orthogonal basis bubble function finite element method (Japanese)
2018/06/19
16:50-18:20 Room #002 (Graduate School of Math. Sci. Bldg.)
Shuji Yoshikawa (Oita University)
Small data global existence for the semi-discrete scheme of a model system of hyperbolic balance laws (Japanese)
Shuji Yoshikawa (Oita University)
Small data global existence for the semi-discrete scheme of a model system of hyperbolic balance laws (Japanese)
2018/05/31
16:30-18:00 Room #056 (Graduate School of Math. Sci. Bldg.)
Olivier Pironneau (Sorbonne University and Academy of Sciences)
Parallel Computing Methods for Quantitative Finance: the Parareal Algorithm for American Options (English)
Olivier Pironneau (Sorbonne University and Academy of Sciences)
Parallel Computing Methods for Quantitative Finance: the Parareal Algorithm for American Options (English)
[ Abstract ]
With parallelism in mind we investigate the parareal method for American contracts both theoretically and numerically. Least-Square Monte-Carlo (LSMC) and parareal time decomposition with two or more levels are used, leading to an efficient parallel implementation which scales linearly with the number of processors and is appropriate to any multiprocessor-memory architecture in its multilevel version. We prove $L^2$ superlinear convergence for an LSMC backward in time computation of American contracts, when the conditional expectations are known, i.e. before Monte-Carlo discretization. In all cases the computing time is increased only by a constant factor, compared to the sequential algorithm on the finest grid, and speed-up is guaranteed when the number of processors is larger than that constant. A numerical implementation will be shown to confirm the theoretical error estimates.
With parallelism in mind we investigate the parareal method for American contracts both theoretically and numerically. Least-Square Monte-Carlo (LSMC) and parareal time decomposition with two or more levels are used, leading to an efficient parallel implementation which scales linearly with the number of processors and is appropriate to any multiprocessor-memory architecture in its multilevel version. We prove $L^2$ superlinear convergence for an LSMC backward in time computation of American contracts, when the conditional expectations are known, i.e. before Monte-Carlo discretization. In all cases the computing time is increased only by a constant factor, compared to the sequential algorithm on the finest grid, and speed-up is guaranteed when the number of processors is larger than that constant. A numerical implementation will be shown to confirm the theoretical error estimates.
2018/05/08
16:50-18:20 Room #002 (Graduate School of Math. Sci. Bldg.)
Norikazu Saito (University of Tokyo)
Various aspects of numerical analysis (Japanese)
Norikazu Saito (University of Tokyo)
Various aspects of numerical analysis (Japanese)
2018/04/17
16:50-18:20 Room #002 (Graduate School of Math. Sci. Bldg.)
Yoshiki Sugitani (Tohoku University)
Introduction to Machine learning and its application to Medical diagnosis (Japanese)
Yoshiki Sugitani (Tohoku University)
Introduction to Machine learning and its application to Medical diagnosis (Japanese)
2018/02/19
15:00-16:00 Room #056 (Graduate School of Math. Sci. Bldg.)
Michael Plum (Karlsruhe Insitute of Technology)
Existence, multiplicity, and orbital stability for travelling waves in a nonlinearly supported beam (English)
Michael Plum (Karlsruhe Insitute of Technology)
Existence, multiplicity, and orbital stability for travelling waves in a nonlinearly supported beam (English)
[ Abstract ]
For a nonlinear beam equation with exponential nonlinearity, we prove existence of at least 36 travelling wave solutions for the specific wave speed c=1.3. Our proof makes heavy use of computer assistance: starting from numerical approximations, we use a fixed point argument to prove existence of solutions "close to" the approximate ones. Furthermore we investigate the orbital stability of these solutions by making use of both analytical and computer-assisted techniques.
For a nonlinear beam equation with exponential nonlinearity, we prove existence of at least 36 travelling wave solutions for the specific wave speed c=1.3. Our proof makes heavy use of computer assistance: starting from numerical approximations, we use a fixed point argument to prove existence of solutions "close to" the approximate ones. Furthermore we investigate the orbital stability of these solutions by making use of both analytical and computer-assisted techniques.
2018/02/19
16:15-17:15 Room #056 (Graduate School of Math. Sci. Bldg.)
Kaori Nagatou (Karlsruhe Insitute of Technology)
An approach to computer-assisted existence proofs for nonlinear space-time fractional parabolic problems (English)
Kaori Nagatou (Karlsruhe Insitute of Technology)
An approach to computer-assisted existence proofs for nonlinear space-time fractional parabolic problems (English)
[ Abstract ]
We consider an initial boundary value problem for a space-time fractional parabolic equation, which includes the fractional Laplacian, i.e. a nonlocal operator. We treat a corresponding local problem which is obtained by the Caffarelli-Silvestre extension technique, and show how to enclose a solution of the extended problem by computer-assisted means.
We consider an initial boundary value problem for a space-time fractional parabolic equation, which includes the fractional Laplacian, i.e. a nonlocal operator. We treat a corresponding local problem which is obtained by the Caffarelli-Silvestre extension technique, and show how to enclose a solution of the extended problem by computer-assisted means.
2017/12/19
16:50-18:20 Room #128 (Graduate School of Math. Sci. Bldg.)
2017/11/28
16:50-18:20 Room #117 (Graduate School of Math. Sci. Bldg.)
Daisuke Koyama (The University of Electro-Communications)
Hybrid discontinuous Galerkin methods for nearly incompressible elasticity problems
(Japanese)
Daisuke Koyama (The University of Electro-Communications)
Hybrid discontinuous Galerkin methods for nearly incompressible elasticity problems
(Japanese)
[ Abstract ]
A Hybrid discontinuous Galerkin (HDG) method for linear elasticity problems has been introduced by Kikuchi et al. [Theor. Appl. Mech. Japan, vol.57, 395--404 (2009)], [RIMS Kokyuroku, vol.1971, 28--46 (2015)]. We consider to seek numerical solutions of the plane strain problem by the HDG method, especially in the case when materials are nearly incompressible, that is, when the first Lam\'e parameter $\lambda$ is large. In this talk, we consider two cases when the HDG method uses a lifting term and does not use it. When the lifting term is used, the method can be free of volumetric locking. On the other hand, when the lifting term is not used, we have to take an interior penalty parameter of order $\lambda$ as $\lambda$ tends to infinity, in order to guarantee the coercivity of the bilinear form. Taking such an interior penalty parameter causes volumetric locking phenomena. We thus conclude that the lifting term is essential for avoiding the volumetric locking in the HDG method.
A Hybrid discontinuous Galerkin (HDG) method for linear elasticity problems has been introduced by Kikuchi et al. [Theor. Appl. Mech. Japan, vol.57, 395--404 (2009)], [RIMS Kokyuroku, vol.1971, 28--46 (2015)]. We consider to seek numerical solutions of the plane strain problem by the HDG method, especially in the case when materials are nearly incompressible, that is, when the first Lam\'e parameter $\lambda$ is large. In this talk, we consider two cases when the HDG method uses a lifting term and does not use it. When the lifting term is used, the method can be free of volumetric locking. On the other hand, when the lifting term is not used, we have to take an interior penalty parameter of order $\lambda$ as $\lambda$ tends to infinity, in order to guarantee the coercivity of the bilinear form. Taking such an interior penalty parameter causes volumetric locking phenomena. We thus conclude that the lifting term is essential for avoiding the volumetric locking in the HDG method.
2017/11/14
16:50-18:20 Room #002 (Graduate School of Math. Sci. Bldg.)
Ai Ishikawa (Kobe University)
Energy-preserving numerical method based on the variational principle and application to unconstrained optimization problems (Japanese)
Ai Ishikawa (Kobe University)
Energy-preserving numerical method based on the variational principle and application to unconstrained optimization problems (Japanese)
2017/10/23
16:00-17:30 Room #056 (Graduate School of Math. Sci. Bldg.)
Christian Klingenberg (Wuerzburg University, Germany)
On the numerical discretization of the Euler equations with a gravitational force and applications in astrophysics (English)
Christian Klingenberg (Wuerzburg University, Germany)
On the numerical discretization of the Euler equations with a gravitational force and applications in astrophysics (English)
[ Abstract ]
We consider astrophysical systems that are modeled by the multidimensional Euler equations with gravity.
First for the homogeneous Euler equations we look at flow in the low Mach number regime. Here for conventional finite volume discretizations one has excessive dissipation in this regime. We identify inconsistent scaling for low Mach numbers of the numerical fux function as the origin of this problem. Based on the Roe solver a technique that allows to correctly represent low Mach number flows with a discretization of the compressible Euler equations is proposed. We analyze properties of this scheme and demonstrate that its limit yields a discretization of the incompressible limit system.
Next for the Euler equations with gravity we seek well-balanced methods. We describe a numerical discretization of the compressible Euler equations with a gravitational potential. A pertinent feature of the solutions to these inhomogeneous equations is the special case of stationary solutions with zero velocity, described by a nonlinear PDE, whose solutions are called hydrostatic equilibria. We present well-balanced methods, for which we can ensure robustness, accuracy and stability, since it satisfies discrete entropy inequalities.
We will then present work in progress where we combine the two methods above.
We consider astrophysical systems that are modeled by the multidimensional Euler equations with gravity.
First for the homogeneous Euler equations we look at flow in the low Mach number regime. Here for conventional finite volume discretizations one has excessive dissipation in this regime. We identify inconsistent scaling for low Mach numbers of the numerical fux function as the origin of this problem. Based on the Roe solver a technique that allows to correctly represent low Mach number flows with a discretization of the compressible Euler equations is proposed. We analyze properties of this scheme and demonstrate that its limit yields a discretization of the incompressible limit system.
Next for the Euler equations with gravity we seek well-balanced methods. We describe a numerical discretization of the compressible Euler equations with a gravitational potential. A pertinent feature of the solutions to these inhomogeneous equations is the special case of stationary solutions with zero velocity, described by a nonlinear PDE, whose solutions are called hydrostatic equilibria. We present well-balanced methods, for which we can ensure robustness, accuracy and stability, since it satisfies discrete entropy inequalities.
We will then present work in progress where we combine the two methods above.
2017/10/10
16:50-18:20 Room #002 (Graduate School of Math. Sci. Bldg.)
Yumiharu Nakano (Tokyo Institute of Technology)
Meshfree collocation methods for linear and fully nonlinear parabolic equations
Yumiharu Nakano (Tokyo Institute of Technology)
Meshfree collocation methods for linear and fully nonlinear parabolic equations
2017/07/04
16:50-18:20 Room #002 (Graduate School of Math. Sci. Bldg.)
Ming-Cheng Shiue (National Chiao Tung University)
Boundary conditions for Limited-Area Models (English)
Ming-Cheng Shiue (National Chiao Tung University)
Boundary conditions for Limited-Area Models (English)
[ Abstract ]
The problem of boundary conditions in a limited domain is recognized an important problem in geophysical fluid dynamics. This is due to that boundary conditions are proposed to have high resolution over a region of interest. The challenges for proposing later boundary conditions are of two types: on the computational side, if the proposed boundary conditions are not appropriate, it is well-known that the error from the lateral boundary can propagate into the computational domain and make a major effect on the numerical solution; on the mathematical side, the negative result of Oliger and Sundstrom that these equations including the inviscid primitive equations and shallow water equations in the multilayer case are not well-posed for any set of local boundary conditions.
In this talk, three-dimensional inviscid primitive equations and (one-layer and two-layer) shallow water equations which have been used in the limited-area numerical weather prediction modelings are considered. Our goals of this work are two folds: one is to propose boundary conditions which are physically suitable. That is, they let waves move freely out of the domain without producing spurious waves; the other is to numerically implement these boundary conditions by proposing suitable numerical methods. Numerical experiments are presented to demonstrate that these proposed boundary conditions and numerical schemes are suitable.
The problem of boundary conditions in a limited domain is recognized an important problem in geophysical fluid dynamics. This is due to that boundary conditions are proposed to have high resolution over a region of interest. The challenges for proposing later boundary conditions are of two types: on the computational side, if the proposed boundary conditions are not appropriate, it is well-known that the error from the lateral boundary can propagate into the computational domain and make a major effect on the numerical solution; on the mathematical side, the negative result of Oliger and Sundstrom that these equations including the inviscid primitive equations and shallow water equations in the multilayer case are not well-posed for any set of local boundary conditions.
In this talk, three-dimensional inviscid primitive equations and (one-layer and two-layer) shallow water equations which have been used in the limited-area numerical weather prediction modelings are considered. Our goals of this work are two folds: one is to propose boundary conditions which are physically suitable. That is, they let waves move freely out of the domain without producing spurious waves; the other is to numerically implement these boundary conditions by proposing suitable numerical methods. Numerical experiments are presented to demonstrate that these proposed boundary conditions and numerical schemes are suitable.
2017/06/13
16:50-18:20 Room #002 (Graduate School of Math. Sci. Bldg.)
Hirofumi Notsu (Kanazawa University)
Numerical analysis of viscoelastic fluid models (Japanese)
Hirofumi Notsu (Kanazawa University)
Numerical analysis of viscoelastic fluid models (Japanese)
[ Abstract ]
Numerical methods for viscoelastic fluid models are studied. In viscoelastic fluid models the stress tensor is often written as a sum of the viscous stress tensor depending linearly on the strain rate tensor and the extra stress tensor for the viscoelastic contribution. In order to describe the viscoelastic contribution another equation for the extra stress tensor is required. In the talk we mainly deal with the Oldroyd-B and the Peterlin models among several proposed viscoelastic fluid models, and present error estimates of finite element schemes based on the method of characteristics. The key issue in the estimates is the treatment of the divergence of the extra stress tensor appearing in the equation for the velocity and the pressure.
Numerical methods for viscoelastic fluid models are studied. In viscoelastic fluid models the stress tensor is often written as a sum of the viscous stress tensor depending linearly on the strain rate tensor and the extra stress tensor for the viscoelastic contribution. In order to describe the viscoelastic contribution another equation for the extra stress tensor is required. In the talk we mainly deal with the Oldroyd-B and the Peterlin models among several proposed viscoelastic fluid models, and present error estimates of finite element schemes based on the method of characteristics. The key issue in the estimates is the treatment of the divergence of the extra stress tensor appearing in the equation for the velocity and the pressure.
2017/04/25
16:50-18:20 Room #002 (Graduate School of Math. Sci. Bldg.)
Koya Sakakibara (University of Tokyo)
Theory and application of the method of fundamental solutions (日本語)
Koya Sakakibara (University of Tokyo)
Theory and application of the method of fundamental solutions (日本語)
2017/04/11
16:50-18:20 Room #002 (Graduate School of Math. Sci. Bldg.)
Shinya Uchiumi (Waseda University)
Some issues in the Lagrange-Galerkin method and solutions: computability, dependence on the viscosity and inflow boundary conditions (日本語)
Shinya Uchiumi (Waseda University)
Some issues in the Lagrange-Galerkin method and solutions: computability, dependence on the viscosity and inflow boundary conditions (日本語)
2017/01/16
16:50-18:20 Room #117 (Graduate School of Math. Sci. Bldg.)
Hideo Kawarada (AMSOK and Chiba University)
Mathematical model for the generation of calcium carbonate scale (日本語)
Hideo Kawarada (AMSOK and Chiba University)
Mathematical model for the generation of calcium carbonate scale (日本語)
2016/11/21
16:50-18:20 Room #002 (Graduate School of Math. Sci. Bldg.)
Sotirios E. Notaris (National and Kapodistrian University of Athens)
Gauss-Kronrod quadrature formulae (English)
Sotirios E. Notaris (National and Kapodistrian University of Athens)
Gauss-Kronrod quadrature formulae (English)
[ Abstract ]
In 1964, the Russian mathematician A.S. Kronrod, in an attempt to estimate practically the error term of the well-known Gauss quadrature formula, presented a new quadrature rule, which since then bears his name. It turns out that the new rule was related to some polynomials that Stieltjes developed some 70 years earlier, through his work on continued fractions and the moment problem. We give an overview of the Gauss-Kronrod quadrature formulae, which are interesting from both the mathematical and the applicable point of view.
The talk will be expository without requiring any previous knowledge of numerical integration.
In 1964, the Russian mathematician A.S. Kronrod, in an attempt to estimate practically the error term of the well-known Gauss quadrature formula, presented a new quadrature rule, which since then bears his name. It turns out that the new rule was related to some polynomials that Stieltjes developed some 70 years earlier, through his work on continued fractions and the moment problem. We give an overview of the Gauss-Kronrod quadrature formulae, which are interesting from both the mathematical and the applicable point of view.
The talk will be expository without requiring any previous knowledge of numerical integration.
2016/10/31
16:50-18:20 Room #002 (Graduate School of Math. Sci. Bldg.)
Shohiro Sho (Osaka University)
Numerical methods of a quantum hydrodynamic model for semiconductor devices (日本語)
Shohiro Sho (Osaka University)
Numerical methods of a quantum hydrodynamic model for semiconductor devices (日本語)
2016/07/11
16:30-18:00 Room #056 (Graduate School of Math. Sci. Bldg.)
Hiroshi Fujiwara (Kyoto University)
Towards fast and reliable numerical computations of the stationary radiative transport equation (日本語)
Hiroshi Fujiwara (Kyoto University)
Towards fast and reliable numerical computations of the stationary radiative transport equation (日本語)
[ Abstract ]
The radiative transport equation (RTE) is a mathematical model of near-infrared light propagation in human tissue, and its analysis is required to develop a new noninvasive monitoring method of our body or brain activities. Since stationary RTE describes light intensity depending on a position and a direction, a discretization model of 3D-RTE is essentially a five dimensional problem. Therefore to establish a reliable and practical numerical method, both theoretical numerical analysis and computing techniques are required.
We firstly introduce huge-scale computation examples of RTE with bio-optical data. A high-accurate numerical cubature on the unit sphere and a hybrid parallel computing technique using GPGPU realize fast computation. Secondly we propose a semi-discrete upwind finite volume method to RTE. We also show its error estimate in two dimensions.
This talk is based on joint works with Prof. Y.Iso, Prof. N.Higashimori, and Prof. N.Oishi (Kyoto University).
The radiative transport equation (RTE) is a mathematical model of near-infrared light propagation in human tissue, and its analysis is required to develop a new noninvasive monitoring method of our body or brain activities. Since stationary RTE describes light intensity depending on a position and a direction, a discretization model of 3D-RTE is essentially a five dimensional problem. Therefore to establish a reliable and practical numerical method, both theoretical numerical analysis and computing techniques are required.
We firstly introduce huge-scale computation examples of RTE with bio-optical data. A high-accurate numerical cubature on the unit sphere and a hybrid parallel computing technique using GPGPU realize fast computation. Secondly we propose a semi-discrete upwind finite volume method to RTE. We also show its error estimate in two dimensions.
This talk is based on joint works with Prof. Y.Iso, Prof. N.Higashimori, and Prof. N.Oishi (Kyoto University).
2016/06/13
16:30-18:00 Room #056 (Graduate School of Math. Sci. Bldg.)
Atsushi Suzuki (Osaka University)
Dissection : A direct solver with kernel detection for finite element matrices
(日本語)
Atsushi Suzuki (Osaka University)
Dissection : A direct solver with kernel detection for finite element matrices
(日本語)
[ Abstract ]
Large-scale sparse matrices are solved in finite element analyses of elasticity and/or flow problems. In some cases, the matrix may be singular, e.g. due to pressure ambiguity of the Navier-Stokes equations, or due to rigid body movements of sub-domain elasticity problems by a domain decomposition method. Therefore, it is better the linear solver understands rank-deficiency of the matrix.
By assuming the matrix is factorized into LDU form with a symmetric partial permutation, and by introducing a threshold to postpone factorization for pseudo null pivots, solvability of the last Schur complement matrix will be examined. Usual procedure for rank-deficiency problem is based on computation of eigenvalues or singular values and an introduced threshold determines the null space. However, developed new algorithm in DOI:10.1002/nme.4729 is based on computation of residuals combined with orthogonal projections onto supposed image spaces and there is no necessary to introduce a threshold for understanding zero value in floating point. The algorithm uses higher precision arithmetic, e.g. quadruple precision, to distinguish numerical round-off errors that occurred during factorization of the whole sparse matrix from ones during the kernel detection procedure itself.
This is joint work with François-Xavier Roux (LJLL, UPMC/ONERA).
Large-scale sparse matrices are solved in finite element analyses of elasticity and/or flow problems. In some cases, the matrix may be singular, e.g. due to pressure ambiguity of the Navier-Stokes equations, or due to rigid body movements of sub-domain elasticity problems by a domain decomposition method. Therefore, it is better the linear solver understands rank-deficiency of the matrix.
By assuming the matrix is factorized into LDU form with a symmetric partial permutation, and by introducing a threshold to postpone factorization for pseudo null pivots, solvability of the last Schur complement matrix will be examined. Usual procedure for rank-deficiency problem is based on computation of eigenvalues or singular values and an introduced threshold determines the null space. However, developed new algorithm in DOI:10.1002/nme.4729 is based on computation of residuals combined with orthogonal projections onto supposed image spaces and there is no necessary to introduce a threshold for understanding zero value in floating point. The algorithm uses higher precision arithmetic, e.g. quadruple precision, to distinguish numerical round-off errors that occurred during factorization of the whole sparse matrix from ones during the kernel detection procedure itself.
This is joint work with François-Xavier Roux (LJLL, UPMC/ONERA).
2016/05/23
16:30-18:00 Room #056 (Graduate School of Math. Sci. Bldg.)
Keiichi Morikuni (University of Tsukuba)
Inner-iteration preconditioning for least squares problems and its applications (日本語)
Keiichi Morikuni (University of Tsukuba)
Inner-iteration preconditioning for least squares problems and its applications (日本語)
[ Abstract ]
We discuss inner-iteration preconditioning for Krylov subspace methods for solving large-scale linear least squares problems. The preconditioning uses several steps of stationary iterative methods, and is efficient when the successive overrelaxation (SOR) method for the normal equations is employed. The SOR inner-iteration left/right-preconditioned generalized minimal residual (BA/AB-GMRES) methods determine a least squares solution/the minimum-norm solution of linear systems of equations without breakdown even in the rank-deficient case. The inner-iteration preconditioning requires less memory than incomplete matrix factorization-type one, and is effective for ill-conditioned and/or rank-deficient least squares problems.
We present applications of inner-iteration preconditioning to solutions of (1) general least squares problems, which is to find a least squares solution whose Euclidean norm is minimum; (2) linear systems of equations which arise in an interior-point method for solving linear programming problems. In (1), we focus on a two-step procedure for the solution of general least squares problems; the first step is to determine a least squares solution and the second step is to determine the minimum-norm solution to a linear system of equation. The solution of each step can be done by using the inner-iteration preconditioned GMRES methods. Numerical experiments show that the SOR inner-iteration preconditioned GMRES methods are more efficient than previous methods for some problems. In (2), the linear systems of equations at each interior-point step become ill-conditioned in the late phase of the interior-point iterations. To solve the linear systems of equation robustly, the inner-iteration preconditioning applies. We present efficient techniques to apply the inner-iteration preconditioning to the linear systems of equations. Numerical experiments on benchmark problems show that the inner-iteration preconditioning is robust compared to previous methods. (2) is joint work with Yiran Cui (University College London), Takashi Tsuchiya (National Graduate Institute for Policy Studies) , and Ken Hayami (National Institute of Informatics and SOKENDAI).
We discuss inner-iteration preconditioning for Krylov subspace methods for solving large-scale linear least squares problems. The preconditioning uses several steps of stationary iterative methods, and is efficient when the successive overrelaxation (SOR) method for the normal equations is employed. The SOR inner-iteration left/right-preconditioned generalized minimal residual (BA/AB-GMRES) methods determine a least squares solution/the minimum-norm solution of linear systems of equations without breakdown even in the rank-deficient case. The inner-iteration preconditioning requires less memory than incomplete matrix factorization-type one, and is effective for ill-conditioned and/or rank-deficient least squares problems.
We present applications of inner-iteration preconditioning to solutions of (1) general least squares problems, which is to find a least squares solution whose Euclidean norm is minimum; (2) linear systems of equations which arise in an interior-point method for solving linear programming problems. In (1), we focus on a two-step procedure for the solution of general least squares problems; the first step is to determine a least squares solution and the second step is to determine the minimum-norm solution to a linear system of equation. The solution of each step can be done by using the inner-iteration preconditioned GMRES methods. Numerical experiments show that the SOR inner-iteration preconditioned GMRES methods are more efficient than previous methods for some problems. In (2), the linear systems of equations at each interior-point step become ill-conditioned in the late phase of the interior-point iterations. To solve the linear systems of equation robustly, the inner-iteration preconditioning applies. We present efficient techniques to apply the inner-iteration preconditioning to the linear systems of equations. Numerical experiments on benchmark problems show that the inner-iteration preconditioning is robust compared to previous methods. (2) is joint work with Yiran Cui (University College London), Takashi Tsuchiya (National Graduate Institute for Policy Studies) , and Ken Hayami (National Institute of Informatics and SOKENDAI).
