Thursday,
April 6, 2000, 3:40 p.m. in PSA Room 103
Kenneth Holmström
Department of Mathematics and Physics, Mälardalen University, Sweden
The TOMLAB Optimization Environment in Matlab
Abstract
TOMLAB is a general purpose, open and integrated Matlab
environment aimed at the solution of a wide range of optimization
problems, as well as for research and teaching in optimization.
One motivation for TOMLAB is to simplify research on applied
optimization problems, giving easy access to all types of solvers;
at the same time having full access to the power of Matlab.
This talk discusses the design and contents of TOMLAB, its performance,
and presents some applications where it has been successfully used.
More than 65 different algorithms for linear, discrete and nonlinear
optimization are implemented.
It is possible to call solvers in the Math Works Optimization Toolbox and
TOMLAB v2.0 also has a call-compatible interface.
We are currently working on a set of predefined MEX-file interfaces
to a large collection of state-of-the-art general-purpose Fortran and C
solvers,
soon to be available.
There are several ways to solve optimization problems in TOMLAB.
Either by a direct call to a solver or using a general multi-solver
driver routine, or interactively, using the Graphical User
Interface (GUI), or a menu system.
The talk will discuss in more detail recent work in
two of the strong areas of TOMLAB,
nonlinear parameter estimation and global optimization.
New algorithms for global optimization of costly functions show
good results on test problems and an industrial train set
optimization problem.
More information on TOMLAB and a demo version for download are
available at the home page of the Applied Optimization and Modeling (TOM)
group, URL: http://www.ima.mdh.se/tom.