Refine
Year of publication
Language
- English (1103) (remove)
Keywords
- optimal control (27)
- stability (14)
- integer programming (11)
- Stochastic programming (9)
- finite elements (9)
- mixed integer programming (9)
- Hamiltonian matrix (8)
- finite element method (8)
- model reduction (8)
- state constraints (8)
A new method for noise removal of arbitrary surfaces
meshes is presented which focuses on the preservation
and sharpening of non-linear geometric features such
as curved surface regions and feature lines. Our method
uses a prescribed mean curvature flow (PMC) for simplicial
surfaces which is based on three new contributions:
1. the definition and efficient calculation of a
discrete shape operator and principal curvature properties
on simplicial surfaces that is fully consistent with
the well-known discrete mean curvature formula, 2. an
anisotropic discrete mean curvature vector that combines
the advantages of the mean curvature normal with
the special anisotropic behaviour along feature lines of
a surface, and 3. an anisotropic prescribed mean curvature
flow which converges to surfaces with an estimated
mean curvature distribution and with preserved nonlinear
features. Additionally, the PMC flow prevents
boundary shrinkage at constrained and free boundary
segments.
We study perturbations of a stochastic program with a probabilistic constraint and r-concave original probability distribution. First we improve our earlier results substantially and provide conditions implying Hölder continuity properties of the solution sets w.r.t. the Kolmogorov distance of probability distributions. Secondly, we derive an upper Lipschitz continuity property for solution sets under more restrictive conditions on the original program and on the perturbed probability measures. The latter analysis applies to linear-quadratic models and is based on work by Bonnans and Shapiro. The stability results are illustrated by numerical tests showing the different asymptotic behaviour of parametric and nonparametric estimates in a program with a normal probabilistic constraint.
We consider stochastic programs with risk measures in the objective and study
stability properties as well as decomposition structures. Thereby we place emphasis on dynamic
models, i.e., multistage stochastic programs with multiperiod risk measures. In this context, we
define the class of polyhedral risk measures such that stochastic programs with risk measures taken
from this class have favorable properties. Polyhedral risk measures are defined as optimal values of
certain linear stochastic programs where the arguments of the risk measure appear on the right-hand
side of the dynamic constraints. Dual representations for polyhedral risk measures are derived and
used to deduce criteria for convexity and coherence. As examples of polyhedral risk measures we
propose multiperiod extensions of the Conditional-Value-at-Risk.
We consider multistage stochastic optimization models containing nonconvex constraints, e.g.,
due to logical or integrality requirements. We study three variants of Lagrangian relaxations and of the corresponding
decomposition schemes, namely, scenario, nodal and geographical decomposition. Based on
convex equivalents for the Lagrangian duals, we compare the duality gaps for these decomposition schemes.
The first main result states that scenario decomposition provides a smaller or equal duality gap than nodal
decomposition. The second group of results concerns large stochastic optimization models with loosely coupled
components. The results provide conditions implying relations between the duality gaps of geographical
decomposition and the duality gaps for scenario and nodal decomposition, respectively.
Portfolio and risk management problems of power
utilities may be modeled by multistage stochastic programs. These
models use a set of scenarios and corresponding probabilities
to model the multivariate random data process (electrical load,
stream flows to hydro units, and fuel and electricity prices). For
most practical problems the optimization problem that contains
all possible scenarios is too large. Due to computational complexity
and to time limitations this program is often approximated by
a model involving a (much) smaller number of scenarios. The proposed
reduction algorithms determine a subset of the initial scenario
set and assign new probabilities to the preserved scenarios.
The scenario tree construction algorithms successively reduce the
number of nodes of a fan of individual scenarios by modifying the
tree structure and by bundling similar scenarios. Numerical experience
is reported for constructing scenario trees for the load
and spot market prices entering a stochastic portfolio management
model of a German utility
We present a mixed-integer multistage stochastic programming model for the short term unit commitment of a hydro-thermal power system under uncertainty in load, inflow to reservoirs, and prices for fuel and delivery contracts. The model is implemented for uncertain load and tested on realistic data from a German power utility. Load scenario trees are generated by a procedure consisting of two steps: (i) Simulation of load scenarios using an explicit respresentation of the load distribution and (ii) construction of a tree out of these scenarios. The dimension of the corresponding mixed-integer programs ranges up to 200,000 binary and 350,000 continuous variables. The model is solved by a Lagrangian-based decomposition strategy exploiting the loose coupling structure. Solving the Lagrangian dual by a proximal bundle method leads to a successive decomposition into single unit subproblems, which are solved by specific algorithms. Finally, Lagrangian heuristics are used to construct nearly optimal first stage decisions.
Mathematical models for the electricity portfolio
management of a utility that owns a hydro-thermal generation system
and trades on the power market often lead to complex stochastic
optimization problems. We present a new approach to solving
stochastic hydro-storage subproblems that occur when stochastic
Lagrangian relaxation is applied to solving such models. The special
structure of such hydro-storage subproblems allows the design
of a stochastic network flow algorithm. The algorithm represents
a stochastic extension of a relaxation method, that algorithmically
solves the linear minimum cost flow problem. It is based on the
iterative improvement of dual costs. Numerical experience of the
new algorithm is reported and its performance is compared with
that of standard LP software .
This work is concerned with transparent boundary conditions (TBCs) for systems of Schrödinger type equations, namely the time-dependent kp-Schrödinger equations. These TBCs
have to be constructed for the discrete scheme, in order to maintain stability and to avoid
numerical re
ections. The discrete transparent boundary conditions (DTBCs) are constructed
using the solution of the exterior problem with Laplace and Z-transformation respectively.
Hence we will analyse the numerical error caused by the inverse Z-transformation. Since
these DTBCs are non-local in time and thus very costly, we present approximate DTBCs,
that allow a fast calculation of the boundary terms.
It is known that for each combinatorial type of convex 3-dimensional
polyhedra, there is a representative with edges tangent to the unit sphere.
This representative is unique up to projective transformations that fix the unit
sphere. We show that there is a unique representative (up to congruence) with
edges tangent to the unit sphere such that the origin is the barycenter of the
points where the edges touch the sphere.
Given a set of service requests (events), a set of guided servers (units),
and a set of unguided service contractors (conts), the vehicle dispatching problem
VDP is the task to find an assignment of events to units and conts as well as tours
for all units starting at their current positions and ending at their home positions
(dispatch) such that the total cost of the dispatch is minimized.
The cost of a dispatch is the sum of unit costs, cont costs, and event costs. Unit
costs consist of driving costs, service costs and overtime costs; cont costs consist of
a fixed cost per service; event costs consist of late costs linear in the late time, which
occur whenever the service of the event starts later than its deadline.
The program ZIBDIP based on dynamic column generation and set partitioning
yields solutions on heavy-load real-world instances (215 events, 95 units) in less
than a minute that are no worse than 1% from optimum on state-of-the-art personal
computers.
We consider a particle constrained to a submanifold ? of the configuration space
Rm. Using that the notion of holonomic constraints coincides with integrability of the corresponding
vector field, we show how this property naturally determines local coordinates on ? . We give a
rigorous justification for the calculation of the mean force along the constrained coordinates, and
we provide a concise geometrical interpretation of the different contributions to the mean force.
Our approach gives rise to a generalisation of the Fixman Theorem which is well known and widely
used in molecular dynamics applications. It further allows for working out a Hybrid Monte-Carlo
based algorithm that can be used to compute arbitrary statistical quantities from constrained
simulations such as the mean force in the context of thermodynamic free energy statistics.
Models for physical systems often take the form of implicit or behavioral models. One important problem is the
identification of which combinations of variables are good candidtates for control variables. This paper first provides one
solution to this problem for linear time varying systems. The solution is shown to be related to a general optimization
problem. It is then shown how these same algorithms can be extended to a large and important class of nonlinear systems.
In this paper the numerical approximation of solutions of Itô stochastic differential
equations is considered, in particular for equations with a small parameter ? in the noise coex-
cient. We construct stochastic linear multi-step methods and develop the fundamental numerical
analysis concerning their mean-square consistency, numerical stability in the mean-square sense and
mean-square convergence. For the special case of two-step Maruyama schemes we derive conditions
guaranteeing their mean-square consistency. Further, for the small noise case we obtain expansions
of the local error in terms of the stepsize and the small parameter ?. Simulation results using several
explicit and implicit stochastic linear k-step schemes, k = 1; 2, illustrate the theoretical findings.
A strategy for controlling the stepsize in the numerical integration of stochastic
differential equations (SDEs) is presented. It is based on estimating the p-th mean of
local errors. The strategy leads to deterministic stepsize sequences that are identical
for all paths. For the family of Euler schemes for SDEs with small noise we derive
computable estimates for the dominating term of the p-th mean of local errors
and show that the strategy becomes efficient for reasonable stepsizes. Numerical
experience is reported for test examples including scalar SDEs and a stochastic
circuit model.
Single-hop WDM networks with a central Passive Star Coupler (PSC), as well as single-hop networks with
a central Arrayed-Waveguide Grating (AWG) and a single transceiver at each node, have been extensively
studied as solutions for the quickly increasing amounts of unicast and multicast traffic in the metropolitan
area. The main bottlenecks of these networks are the lack of spatial wavelength reuse in the studied PSC
based networks and the single transceiver in the studied AWG based metro WDM networks. In this paper
we develop and evaluate the FT EE ???? FREE AWG network, which is based on a central AWG and has arrays
of fixed-tuned transmitters and receivers at each node. Transceiver arrays are a mature technology, making
the proposed network practical. In addition, the transmitter arrays allow for high speed signaling over the
AWG while the receiver arrays relieve the receiver bottleneck arising from multicasting in conjunction with
spatial wavelength reuse on the AWG. Our results from probabilistic analysis and simulation indicate that
the FTEE ???? FREE AWG network gives particularly good throughput-delay performance for a mix of unicast
and multicast traffic.
Metastability in reversible diffusion processes I. Sharp asymptotics for capcities and exit times
(2004)
We develop a potential theoretic approach to the problem of metastability for reversible diffusion processes with generators of the form +rF ( )r on R or subsets of , where F is a smooth function with finitely many local minima. In analogy to previous work in discrete Markov chains, we show that metastable exit times from the attractive domains of the minima of F can be related, up to multiplicative errors that tend to one as # 0, to the capacities of suitably constructed sets. We show that this capacities can be computed, again up to multiplicative errors that tend to one, in terms of local characteristics of F at the starting minimum and the relevant saddle points. As a result, we are able to give the first rigorous proof of the classical Eyring-Kramers formula in dimension larger than 1. The estimates on capacities make use of their variational representation and monotonicity properties of Dirichlet forms. The methods developed here are extensions of our earlier work on discrete Markov chains to continuous diffusion processes.
We continue the analysis of the problem of metastability for reversible diffusion processes,
initiated in [BEGK3], with a precise analysis of the low-lying spectrum of the generator.
Recall that we are considering processes with generators of the form 1+rF()r on Rd or subsets
of Rd , where F is a smooth function with finitely many local minima. Here we consider only
the generic situation where the depths of all local minima are different. We show that in general
the exponentially small part of the spectrum is given, up to multiplicative errors tending to one, by
the eigenvalues of the classical capacity matrix of the array of capacitors made of balls of radius
centered at the positions of the local minima of F. We also get very precise uniform control on the
corresponding eigenfunctions. Moreover, these eigenvalues can be identified with the same precision
with the inverse mean metastable exit times from each minimum. In [BEGK3] it was proven
that these mean times are given, again up to multiplicative errors that tend to one, by the classical
Eyring–Kramers formula.
In this survey, we show that various stochastic optimization problems arising in
option theory, in dynamical allocation problems, and in the microeconomic theory
of intertemporal consumption choice can all be reduced to the same problem of
representing a given stochastic process in terms of running maxima of another
process. We describe recent results of Bank and El Karoui (2002) on the general
stochastic representation problem, derive results in closed form for Lévy processes
and diffusions, present an algorithm for explicit computations, and discuss some
applications.
This paper is concerned with the effficient implementation of transparent boundary conditions
(TBCs) for wide angle parabolic equations (WAPEs) assuming cylindrical symmetry.
In [1] a discrete TBC of convolution type was derived from the fully discretized whole?space
problem that is reflection?free and yields an unconditionally stable scheme. Since the discrete
TBC includes a convolution with respect to range with a weakly decaying kernel, its
numerical evaluation becomes very costly for long-range simulations.
As a remedy we construct new approximative transparent boundary conditions involving
exponential sums as an approximation to the convolution kernel. This special approximation
enables us to use a fast evaluation of the convolution type boundary condition.
This new approach was outlined in detail in [2] for the standard "parabolic" equation.
Differential algebraic equations with properly stated leading term are equations of the form A(x(t),t)(d(x(t),t))'+b(x(t),t)=0 with in some sense well-matched coefficients. Systems resulting from the modified nodal analysis (MNA) in circuit simulation promptly fit into this form. Recent results concerning solvability and numerical treatment of those equations are discussed. An index notion that works via linearization is given. This allows for index criteria just in terms of the coefficients A,d,b and their first partial derivatives, no further derivative arrays are used.
We present a way to efficiently treat the well-known transparent boundary
conditions for the Schrödinger equation. Our approach is based on two ideas:
firstly, to derive a discrete transparent boundary condition (DTBC) based on the
Crank-Nicolson finite difference scheme for the governing equation. And, secondly,
to approximate the discrete convolution kernel of DTBC by sum-of-exponentials for
a rapid recursive calculation of the convolution. We illustrate the efficiency of the
proposed method on several examples.
The index of DAE systems arising from linear quadratic optimal control problems is considered. Necessary and sufficient conditions ensuring regularity with tractability index one are proved. Then, it is shown that if the control problem DAE is regular with index one and if the leading term of the DAE to be controlled is given by one full-column-rank and one full-row-rank matrix, then it has a Hamiltonian inherent explicit ODE. For problems with regular index zero or index one DAEs to be controlled, the DAE of the control problem is shown to be regular with tractability index one or three, depending on whether the control coefficient R is singular.
By the use of the corresponding shift matrix, the paper gives a criterion for the unique solvability of linear boundary value problems posed for linear differential algebraic equations up to index 2 with well-matched leading coefficients. The solution is constructed by a proper Green function. Another characterization of the solutions is based upon the description of arbitrary affine linear subspaces of solutions to linear differential algebraic equations in terms of solutions to the adjoint equation. When applied to boundary value problems, the result provides a constructive criterion for unique solvability and allows reducing the problem to initial value problems and linear algebraic equations.
We consider a particle constrained to a submanifold ? of the
configuration space Rm. Using that the notion of holonomic constraints coincides
with integrability of the corresponding vector field, we show how this property
naturally determines local coordinates on ?. We give a rigorous justification for
the calculation of the mean force along a constrained coordinate, and we provide
a concise geometrical interpretation of the different contributions to the mean
force in terms of the unconstrained vector field and extrinsic curvature properties
of ? in Rm. Our approach gives rise to a Hybrid Monte-Carlo based algorithm
that can be used to compute the mean force acting on selected coordinates in the
context of thermodynamic free energy statistics.
We analyze an interactive model of credit ratings where external shocks, initially
affecting only a small number of firms, spread by a contagious chain reaction to the
entire economy. Counterparty relationships along with discrete adjustments of credit
ratings generate a transition mechanism that allows the financial distress of one firm
to spill over to its business partners. Such a contagious infectious of financial distress
constitutes a source of intrinsic risk for large portfolios of credit sensitive securities that
cannot be “diversified away.” We provide a complete characterization of the fluctuations
of credit ratings in large economies when adjustments follow a threshold rule. We also
analyze the effects of downgrading cascades on aggregate losses of credit portfolios. We
show that the loss distribution has a power-law tail if the interaction between different
companies is strong enough.
Stability of Linear Stochastic Difference Equations in Strategically Controlled Random Environments
(2004)
We consider the stochastic sequence fYtgt2N defined recursively by the linear relation
Yt+1 = AtYt+Bt in a random environment. The environment is described by the stochastic
process f(At;Bt)gt2N and is under the simultaneous control of several agents playing a
discounted stochastic game. We formulate sufficient conditions on the game which ensure
the existence of Nash equilibrium in Markov strategies which has the additional property
that, in equilibrium, the process fYtgt2N converges in distribution to a stationary regime.
We study the effect of investor inertia on stock price fluctuations with a market microstructure
model comprising many small investors who are inactive most of the time.
It turns out that semi-Markov processes are tailor made for modelling inert investors.
With a suitable scaling, we show that when the price is driven by the market imbalance,
the log price process is approximated by a process with long range dependence
and non-Gaussian returns distributions, driven by a fractional Brownian motion. Consequently,
investor inertia may lead to arbitrage opportunities for sophisticated market
participants. The mathematical contributions are a functional central limit theorem for
stationary semi-Markov processes, and approximation results for stochastic integrals
of continuous semimartingales with respect to fractional Brownian motion.
We consider a financial market model with a large number of interacting
agents. Investors are heterogeneous in their expectations
about the future evolution of an asset price process. Their current
expectation is based on the previous states of their “neighbors” and
on a random signal about the “mood of the market.” We analyze the
asymptotics of both aggregate behavior and asset prices. We give sufficient
conditions for the distribution of equilibrium prices to converge to
a unique equilibrium, and provide a microeconomic foundation for the
use of diffusion models in the analysis of financial price fluctuations.
We consider general economies in which rational agents interact locally. The local aspect
of the interactions is designed to represent in a simple abstract way social interactions, that
is, socioeconomic environments in which markets do not mediate all of agents' choices, and
each agent's choice might be in part determined, for instance, by family, peer group, or ethnic
group effects. We study static as well as dynamic infinite horizon economies; we allow for
economies with incomplete information, and we consider jointly global and local interactions,
to integrate e.g., global externalities and markets with peer and group effects. We provide
conditions under which such economies have rational expectations equilibria.
We illustrate the effects of local interactions when agents are rational by studying in detail
the equilibrium properties of a simple economy with quadratic preferences which captures, in
turn, local preferences for conformity, habit persistence, and preferences for status or adherence
to aggregate norms of behavior.
We give sufficient conditions for a non-zero sum discounted stochastic game with
compact and convex action spaces and with norm-continuous transition probabilities,
but with possibly unbounded state space, to have a Nash equilibrium in homogeneous
Markov strategies that depends in a Lipschitz continuous manner on the current state. If
the underlying state space is compact this yields the existence of a stationary equilibrium.
Stochastic games with weakly interacting players provide a probabilistic framework within
which to study strategic behavior in models of non-market interactions.
This paper addresses the regularization of pointwise state constraints in optimal
control problems. By analyzing the associated dual problem, it is shown that the regularized problems
admit Lagrange multipliers in L2-spaces. Under a certain boundedness assumption, the solution of
the regularized problem converges to the one of the original state constrained problem. The results
of our analysis are confirmed by numerical tests.
An optimal control problem for a 2-d elliptic equation is investigated with pointwise control constraints.
This paper is concerned with discretization of the control by piecewise constant functions. The state and
the adjoint state are discretized by linear finite elements. Approximations of the optimal solution of the continuous
optimal control problem will be constructed by a projection of the discrete adjoint state. It is proved that these
approximations have convergence order h2.
We provide a number of new construction techniques for cubical complexes and cubical
polytopes, and thus for cubifications (hexahedral mesh generation). As an application we
obtain an instance of a cubical 4-polytope that has a non-orientable dual manifold (a Klein
bottle). This confirms an existence conjecture of Hetyei (1995).
More systematically, we prove that every normal crossing codimension one immersion of
a compact 2-manifold into R3 is PL-equivalent to a dual manifold immersion of a cubical
4-polytope. As an instance we obtain a cubical 4-polytope with a cubation of Boy's surface
as a dual manifold immersion, and with an odd number of facets. Our explicit example has
17 718 vertices and 16 533 facets. Thus we get a parity changing operation for 3-dimensional
cubical complexes (hexa meshes); this solves problems of Eppstein, Thurston, and others.
We discuss solvers for Sylvester, Lyapunov, and Stein equations
that are available in the SLICOT Library (Subroutine
Library In COntrol Theory). These solvers offer improved
efficiency, reliability, and functionality compared to corresponding
solvers in other computer-aided control system design
packages. The performance of the SLICOT solvers is
compared with the corresponding MATLAB solvers.
The paper presents a unified approach to local likelihood estimation
for a broad class of nonparametric models, including e.g. the regression,
density, Poisson and binary response model. The method extends
the adaptive weights smoothing (AWS) procedure introduced in Polzehl
and Spokoiny (2000) in context of image denoising. Performance of the
proposed procedure is illustrated by a number of numerical examples
and applications to density or volatility estimation, classification and
estimation of the tail index parameter. We also establish a number of
important theoretical results on properties of the proposed procedure.
The adaptive weights smoothing (AWS) procedure was introduced in
Polzehl and Spokoiny (2000) in the context of image denoising. The
procedure has some remarkable properties like preservation of edges and
contrast, and (in some sense) optimal reduction of noise. The procedure
is fully adaptive and dimension free. Simulations with artificial images
show that AWS is superior to classical smoothing techniques especially
when the underlying image function is discontinuous and can be well
approximated by a piecewise constant function. However, the latter as-
sumption can be rather restrictive for a number of potential applications.
Here the AWS method is generalized to the case of an arbitrary local lin-
ear parametric structure. We also establish some important results about
properties of the AWS procedure including the so called "propagation
condition" and spatial adaptivity. The performance of the procedure is
illustrated by examples for local polynomial regression in univariate and
bivariate situations.
Error estimates for the numerical approximation of boundary semilinear elliptic control problems
(2004)
We study the numerical approximation of boundary optimal control problems governed
by semilinear elliptic partial differential equations with pointwise constraints on the control.
The analysis of the approximate control problems is carried out. The uniform convergence of discretized
controls to optimal controls is proven under natural assumptions by taking piecewise constant
controls. Finally, error estimates are established.
Regular Lagrange multipliers for control problems with mixed pointwise control-state constraints
(2004)
A class of quadratic optimization problems in Hilbert spaces is considered, where
pointwise box constraints and constraints of bottleneck type are given. The main focus is to prove the
existence of regular Lagrange multipliers in L2-spaces. This question is solved by investigating the
solvability of a Lagrange dual quadratic problem. The theory is applied to different optimal control
problems for elliptic and parabolic partial differential equations with mixed pointwise control-state
constraints.
Let Ex be a collection of i.i.d. exponential random
variables. Symmetric Bouchaud’s model on Z2 is a Markov chain
X(t) whose transition rates are given by wxy = ν exp(−βEx ) if x,
y are neighbours in Z2 . We study the behaviour of two correlation functions: P[X(tw + t) = X(tw )] and P X(t ) = X(tw )∀t ∈
[tw , tw + t] . We prove the (sub)aging behaviour of these functions
when β > 1.
A popular model to describe credit risk in practice is CreditRisk
+
and in
this paper a Fourier inversion to obtain the distribution of the credit loss is
proposed. A deeper analysis of the Fourier transformation showed that there
are at least two methods to obtain the distribution although the corresponding
characteristic function is not integrable.
The CreditRisk
+
model will be extended such, that general dependent sec-
tor variables can be taken into consideration, for example dependent lognormal
sector variables. Then the transfer to a continuous time model will be per-
formed and the sector variables become processes, more precisely geometric
Brownian motions.
To have a time continuous credit risk model is an important step to combine
this model with market risk. Additionally a portfolio model will be presented
where the changes of the spreads are driven by the sector variables. Using a
linear expansion of the market risk, the distribution of this portfolio can be
determined. In the special case that there is no credit risk, this model yields
the well known Delta normal approach for market risk, hence a link between
credit risk and market risk has been established.
The CreditRisk model launched by CSFB in 1997 is widely used by practitioners in the banking sector as a simple means for the quantification of credit
risk, primarily of the loan book. We present an alternative numerical recursion scheme for CreditRisk, equivalent to an algorithm recently proposed by
Giese, based on well-known expansions of the logarithm and the exponential
of a power series. We show that it is advantageous to the Panjer recursion
advocated in the original CreditRisk
document, in that it is numerically stable. The crucial stability arguments are explained in detail. Furthermore, the
computational complexity of the resulting algorithm is stated.
We introduce a new Monte Carlo method for constructing the exercise
boundary of an American option in a generalized Black-Scholes framework.
Based on a known exercise boundary, it is shown how to price and hedge the
American option by Monte Carlo simulation of suitable probabilistic represen-
tations in connection with the respective parabolic boundary value problem.
The methods presented are supported by numerical simulation experiments.
In this paper we investigate the use of parallel computing to deal with the high computational cost of numerical algorithms for model reduction of large linear descriptor systems. The state-space truncation methods considered here are composed of iterative schemes which can be efficiently implemented on parallel architectures using existing parallel linear algebra libraries. Our experimental results on a cluster of Intel Pentium processors show the performance of the parallel algorithms.
We discuss a parallel library of efficient algorithms for model reduction of largescale
systems with state-space dimension up to O(104). We survey the numerical
algorithms underlying the implementation of the chosen model reduction methods.
The approach considered here is based on state-space truncation of the system
matrices and includes absolute and relative error methods for both stable and unstable
systems. In contrast to serial implementations of these methods, we employ
Newton-type iterative algorithms for the solution of the major computational tasks.
Experimental results report the numerical accuracy and the parallel performance of
our approach on a cluster of Intel Pentium II processors.
We describe a prototype web service for model reduction of very large-scale linear systems, with
dimension in the order of millions of states, that includes a user-friendly interface designed so that the computation
can be easily performed via the HTTP protocol. Access via a web browser isolates the user of the service from the
complexities of installing and using the parallel model reduction codes and the maintenance of the hardware. In case
the routines are found to be appropriate for the problem the user needs to solve, the library can be then downloaded
and installed on the user’s own computing resources.
This paper illustrates the major issues of the access procedure by means of graphical examples, and describes
the structure and implementation of the remote model reduction service. The service is offered in a cluster of Linux
machines.
A Structure-Preserving Method for Generalized Algebraic RiccatiEquations Based on Pencil Arithmetic
(2004)
This paper describes a numerical method for extracting the stable
right deflating subspace of a matrix pencil Z Y using
a spectral projection method. It has several advantages compared
to other spectral projection methods like the sign function
method. In particular it avoids the rounding error induced
loss of accuracy associated with matrix inversions. The new algorithm
is particularly well adapted to solving continuous-time
algebraic Riccati equations. In numerical examples, it solves
Riccati equations to high accuracy.