Software
Preface
This
section collects all the software produced o
coauthored by members of the project
research units.
The software available through the section
is largely described in papers published
with the project support: please, visite the
related Papers section.
Available software
- GKLS
generator
Authors:
M. Gaviano, D.E. Kvasov, D. Lera, Ya.D. Sergeyev
Topic:
Software for generation of classes of
test functions with known local and
global minima for global optimization (published in
M. Gaviano, D.E. Kvasov, D. Lera, Ya.D. Sergeyev,
Algorithm 829: Software for generation of classes of test functions
with known local and global minima for global optimization, ACM
Transactions on Mathematical Software, 29(4), 2003, pp. 469-480)
Description:
Development of numerical algorithms for
global optimization is strongly
connected to the problem of construction
of test functions for studying and
verifying validity of these algorithms.
Many of global optimization tests are
taken from real-life applications and
for this reason a complete
information about them is not available.
It often happens that there are not
known a priori the number of local
minima present in the problem, their
locations, regions of attraction, and
even values (including that one of
the global minimum).
The GKLS generator is a procedure for
generating three types (non-differentiable,
continuously differentiable and twice
continuously differentiable) of
classes of test functions with known
local and global minima for
multiextremal multidimensional
box-constrained global optimization.
- TESTVIPs
- A Matlab collection of Variational
Inequality Problems
Authors:
F. Tinti
Topic:
The MatLab files implementing several
variational inequality problems and
nonlinear complementarity problems
arising from the literature are given.
Description:
We consider the following test problems:
- Mathiesen's problem
- Kojima Shindo's problem
- Harker's Nash-Cournot problem
- Pang & Murphy's Nash-Cournot
problem
- Braess Network problem
- User OPT problem
- HpHard problem.
Each VIP or NCP is implemented in
two files: the first contains the
parameters and the definition of the
feasible region of the problem and
the second is the M-function file to
compute the value of the map related
to the problem.
- MECVIPs
- Matlab codes of extragradient methods for VIPs
Authors:
F. Tinti
Topic:
MatLab codes implementing
extragradient-type methods for
pseudomonotone variational inequality
problems.
Description:
We give several MatLab scripts
implementing some extragradient-type
methods. In particular the considered
methods are:
- Extragradient method with
constant steplength;
- Extragradient method with
adaptive steplegth (Marcotte's
version);
- Extragradient method with
adaptive steplegth (first
modified version);
- Extragradient method with
adaptive steplegth (second
modified version);
- Projection-Contraction method (Solodov-Tseng);
- Hyperplane projection method (Solodov-Svaiter).
These codes can be evaluated using
the MatLab test problems collected
in TESTVIPs. The Matlab
implementation of the above methods
enables us to easily change the
parameters, pointing out the
effectiveness and the features of
each algorithm.
- GPDT
- Gradient Projection-based
Decomposition Technique
[Last updated: October 2006]
Authors:
T. Serafini, L. Zanni, G. Zanghirati
Topic:
Software for convex quadratic programs
arising in training the learning
methodology Support Vector Machines
- Hessian matrix: dense and large (bigger
than 104)
- Constraints: box and a single linear
equality constraint
Description:
decomposition technique based on a
sequence of QP subproblems of smaller
size (about 103).
Two Gradient projection methods for the
smaller QP subproblems are available:
- GVPM (Generalized Variable Projection
Method)
linesearch strategy:
limited minimization rule
steplength selection:
adaptive switch between the two
Barzilai-Borwein rules
- DFGPM (Dai-Fletcher Nonmonotone
Gradient Projection Method)
linesearch strategy:
nonmonotone linesearch
steplength selection:
modified Barzilai-Borwein rule
Code:
- serial: standard C++
- parallel: standard
C++ and MPI communication routines
Tested platform
- serial: Intel-based
PC, Digital-Alpha and HP-Itanium2
workstations
- parallel: Cray T3E,
IBM Xeon-based Linux cluster system, IBM
Power5-based AIX cluster.
Benchmark test problems:
- MNIST database of
handwritten digits
- UCI Adult database
income prediction
- Text categorization
- BLKFCLT
- Regularized Cholesky-like
factorization for sparse symmetric matrices
[Last updated: October 2005]
Authors:
S. Bonettini, V. Ruggiero
Topic:
Collection of Fortran 90 subroutines for
the factorization and the solution of
systems that admit a Cholesky-like
factorization. The
code is based on a modification of the
package by Ng and Peyton (included in
the LIPSOL 0.3 package), that performs
the Cholesky factorization of a positive
definite matrix.
IP-PCG
- An interior point algorithm with the PCG method as inner
solver [Last updated: December 2006]
Authors:
S. Bonettini, V. Ruggiero
Topic:
Software for nonlinear constrained optimization.
- Hessian matrix: sparse and large
- Constraints: box, equality and inequality.
Description:
Inexact Newton interior--point algorithm with a line search
strategy based on the Eisenstat and Walker rule.
Code:
- standard C++
- partial AMPL interface
Platform:
- HP Itanium2 workstation.
A
Collection of optimal control problems
[Last updated: December 2006]
Authors:
S. Bonettini, V. Ruggiero
Topic:
A collection of 25 nonlinear and quadratic programming
problems arising from elliptic and parabolic control
problems.
Description:
For each problem it is available the AMPL model; a detailed
description of the discretization technique and of the
structure of the Jacobian and Hessian matrices can be found
in the documentation.
VIsemiLSQRnew2
[Last updated: December 2006]
Authors:
V. Ruggiero, F. Tinti
Topic:
A Fortran 90 software designed to solve large scale variational
inequality problems using the generalisation of the Inexact Newton
method applied to a semismooth nonlinear system..
Description:
Semismooth algorithm with a line search strategy.
We have used LSQR method, combined by a convenient preconditioner
(a variant of Incomplete LU-factorization).
Furthemore we have introduced an adaptive stopping rule.
Code:
- Fortran 90
- partial AMPL interface
Tested platform
- DEC ALPHA 21264 EV6/7 633Mhz workstation
Upcoming software
- A Parallel Matlab Linux
cluster-enabled tool for Dynamic
Magnetic Resonance Imaging
Authors:
G. Landi, E. Loli Piccolomini, F. Zama
Topic:
A parallel Matlab-based software,
implementing a domain decomposition
technique, for dynamic Magnetic
Resonance (MR) sequences reconstruction.
The software is designed for MPI-based
PC Linux clusters running open source
software for the network.
|
|
