|
Arbeitsgruppe
Numerische Mathematik
Sekretariat
Mitglieder
Forschungschwerpunkt
Lehrveranstaltungen
Projekte
Studienverlauf + Diplomarbeiten
|
Scientific work:
-
numerical approximation
-
error analysis
-
numerical linear algebra
-
numerical methods for ODE
-
continuous optimization
Present research:
-
numerical methods for large scale optimization
-
software development
|
|
A very detailed decision-tree for optimization-software with further
links to other sources. Created and maintained by H.D. Mittelmann and P. Spellucci.
|
Available software:
|
Testenvironment for nonlinear programming codes.
|
|
testenviron.tar.gz:
|
A testenvironment for nonlinear programmming codes. Function codes and
drivers
for a collection of currently 361 testcases comprising all of
Hock&Schittkowski, Schittkowski's second collection, Dembo's
collection
and from Himmelblau, augmented by some others. With an easy to use
command-script
for UNIX-affectionates. (480 kB)
|
|
|
DONLP2: SQP/ECQP-method for general continuous nonlinear programming.
|
|
donlp2.tar.gz:
|
Spellucci's mixed SQP/ECQP-method for general continuous nonlinear
programming
problems. This version allows a choice between exact and numerical
gradients. Three
modes of numerical differentiation are provided. It is also now possible
to link with an external softwaresystem for function evaluation (see the
new documentation: donlp2doc.ps). With complete documentation, papers
concerning the underlying theory, shell-script, makefile and
examples. (f77-version) (300 kB)
|
|
donlp2_ansi_c.tar.gz:
|
A ANSI-C version of donlp2, which is functionally completely identical to
the f77-version. (250 kB)
|
|
donlp2_autodiff_cut.tar.gz:
|
A version of donlp2 (in f77) using automatic differentiation to obtain
gradients.
Interface written by H.D. Mittelmann, Tempe. You must add
jakef.f and
support.f
from netlib/jakef to run it.
(137 kB)
|
|
|
DONLP2 and AMPL.
|
|
|
Free student-binary for the modelling language AMPL with interface and
binary for DONLP2.
|
|
README
|
A short introduction.
|
|
donlp2_ampl.ps.gz
|
A users guide for DONPL2 in combination with the AMPL interface (gziped
Postscript 66 kB).
|
|
|
The following binaries are available yet:
|
|
Win9x/NT4
|
For Intel PC under Windows 95/98 and NT4.
|
|
Linux/Intel
|
For Intel PC unter Linux.
|
|
Linux/Alpha
|
For Alpha PC under Linux.
|
|
Unix DEC Alpha
|
For DEC Alpha under DEC-Unix 4.0.
|
|
PowerPC/AIX
|
For PowerPC under AIX 4.3.
|
|
HP/HPUX
|
For HP PA2.0 under HPUX 10.20.
|
|
IBM/AIX
|
For IBM RS6000 under AIX 3.2.
|
|
SGI/IRIX
|
For SGI under IRIX-6.* (32/64 Bit).
|
|
SUN/Solaris
|
For Sun under Solaris 2.*/SunOS 5.*.
|
|
|
Models
|
Some nonlinear models of well known nonlinear optimization problems coded
in AMPL.
|
|
|
Large problems may be solved with a commercial AMPL binary and a DONLP2
binary compiled for the necessary dimensions or by submission to the
NEOS server.
This submission may be done in three different ways: per e-mail, through
the web, and with a special submission tool. In the directory large are
some files for larger problems submitted per e-mail and the resulting
output.
|
|
|
DOMIN: BFGS-method for unconstrained optimization.
|
|
domin.tar.gz:
|
Spellucci's implementation of the BFGS-method for unconstrained
optimization.
With driver, testcases and user-documentation, written in f77 with
milstd1753-extensions. (14 kB)
|
|
domin_c.tar.gz:
|
A f2c' ed version of domin. To use this one you need the libraries
accompanying f2c and a c-compiler allowing ANSI-c.
(gcc will work, e.g.). The remarks concerning use of the
c-version apply of course. (20 kB)
|
|
|
DUALQP: Goldfarb-Idnani-dual QP solver.
|
|
dualqp.tar.gz:
|
Spellucci's implementation of the Goldfarb-Idnani-dual QP solver. With
driver,
examples, and user-documentation. (dense linear algebra). Written in
f77
with milstd1753-extensions. (11 kB)
|
|
dualqp_c.tar.gz:
|
A f2c' ed version of dualqp. To use this one you need the libraries
accompanying f2c and a c-compiler allowing ANSI-c.
(gcc will work, e.g.). The remarks concerning use of the
c-version apply of course. (18 kB)
|
|
|
GPCG: Gradient-Projection/CG-algorithm for bound constrained QP problems.
|
|
gpcg.tar.gz:
|
Rolf Felkels
implementation of
the Moré-Toraldo gradient-projection/cg-algorithm for bound
constrained
QP-problems. f77 (23 kB)
|
|
gpcg_c.tar.gz:
|
Rolf Felkels
implementation of
the Moré-Toraldo gradient-projection/cg-algorithm for bound
constrained
qp-problems. ANSI-c version. (29 kB)
|
|
|
PL2: Gradient-Projection/Lanczos-algorithm for bound constrained nonconvex
problems.
|
|
pl2.tar.gz:
|
Rolf Felkels
implementation of
a bound constrained optimization technique using
second order information f77 Version (150
kB). See also Preprint
Nr. 1995.
|
Selection tool for nonlinear programming problems
Available reports:
|
qp.ps.gz:
|
Paper on solving general convex QP's via a bound constraint exact
quadratic
augmented Lagrangian (revised 6/96). (91 kB compressed postscript)
|
|
detail.ps.gz:
|
Paper with complete list of donlp2's results obtained with the
testenvironment
from above (testenviron.tar.gz). (59 kB compressed postscript)
|
|
secqp.ps.gz:
|
Paper on "A SQP method for general nonlinear programs using only equality
constrained subproblems" (revised 3/96). Paper constitutes part of theory
underlying donlp2. (130 kB compressed postscript)
|
|
newsqpsh.ps.gz:
|
Paper on "A new technique for inconsistent QP problems in the SQP method"
(revised 3/96). Paper constitutes the second part of the theory underlying
donlp2. (130 kB compressed postscript)
|
|
large_scale_qp.ps.gz:
|
Numerical experiments with four modern methods for large scale
QP-problems.
Numerical study of methods for solving convex QP-problems with equality
constraints and bounds on the variables. (345 kB compressed postscript)
|
|
sr1.ps:
|
A modified rank one update which converges Q-superlinearly.
Proof and numerical verification of a well defined modification of
SR1. (255 kB postscript)
|
|
genqr.ps.gz:
|
A QR-decomposition for matrix pencils. (Joint paper with W. Hartmann from
SAS)
(56 kB compressed postscript)
|
|
nloaspekte.ps:
|
Einige Aspekte bei der Beurteilung und Auswahl von Optimierungsverfahren
für kontinuierliche Probleme (Vortragsfolien). Some aspects for
judging and
choosing a method for nonlinear continuous optimization. Foils, written in
german.
(202 kB postscript)
|
|
largescalenlo.ps.gz:
|
Nichtlineare Optimierungsprobleme hoher Dimension: Stand der
Forschung. Large scale nonlinear optimization. State of the art 1998.
Written in german. (198 kB compressed postscript)
|
|
trsqp.ps.gz:
|
Paper on "Numerical experiments with an exact penalty function for convex
inequality constrained QP-problems". Preprint Nr. 1815, TU Darmstadt,
Department
of Mathematics, March 1996. (67 kB compressed postscript)
|
|
stateoftheartn
lo.ps.gz:
|
Nonlinear (local) optimization: The state of the art (2001) (214 kB
compressed Postscript)
|
Teaching:
|
|
In alternation with other colleagues the following lectures are given
regularly: (lecture notes are in German)
|
|
IntroNuMa
|
Introduction to numerical analysis and numerical linear algebra
(Course in WS 2004/05). Contains: Interpolation and
approximation, quadrature, linear and nonlinear systems.
(1.2 MB pdf, last updated 9/2005)
|
|
IntroNuMa2
|
Introduction to Numerical Analysis II. (Revision summer 2005)
Contains: Eigenvalue computations. Iterative solution of linear
systems. Roundoff analysis. Basics of the fast
Fourier transform.
|
|
ANumAnaI
|
Advanced numerical analysis I. Contains: Initial and boundary value
problems of ODE's; elliptic boundary value problems.
(4.2 MB PDF, last updated 02/2006)
|
|
ANumAnaII
|
Advanced numerical analysis II. Time dependent partial differential
equations. (1400 kB PDF, last updated 9/2004)
|
|
NumIng
|
Numerical analysis for engineers and scientists. Contains: linear and
nonlinear
systems, eigenvalues, Interpolation, quadrature, ODE's, PDE's. (last updated 10/2005, 6.0
MB pdf)
|
|
NumAlg
|
Numerical algorithms for computer science. Contains: Polynomials,
interpolation, linear and nonlinear
systems, error analysis. (527 kB compressed postscript, last
updated April 2001)
|
|
NumOpti
|
An introductary text, also suitable for an Engineering optimization
course.
Introduction to linear and nonlinear optimization. Contains: Theory and
computations. (1.1 MB pdf, last updated March 2006)
Teaching textbook: Numerische Verfahren der nichtlinearen Optimierung
(Numerical methods of nonlinear optimization. Written in german)
Birkhäuser, Basel 1993. ISBN 3-7643-2854-1.
|
|
Opti
|
Intended for an advanced mathematics course.
Modern methods of convex optimization (746 kB compressed postscript, last updated July
2000)
|
|
NUMMB
|
Introduction to numerical analysis for mechanical engineers and mechanics
(pdf, last updated 4/2005)
|
|
