Monday, May 20
1:30-3:30 PM
Salon B
MS9
Parameter Estimation and Optimum Experimental Design
Mathematical models for real life processes typically contain parameters that have to be
determined by experiment. The speakers will examine two issues related to this process: The
design of experiments that produce high quality data for parameter estimation ("optimum
experimental design") and the task of actually estimating the parameters from given experimental
data ("parameter estimation"). Their presentations describe recent developments in numerical
methods for the treatment of complex nonlinear models, especially DAE and PDE boundary value
problems. They will discuss reduced Gauss Newton and SQP type methods, stepsize strategies,
exploitation of structures, parallel algorithms, and applications from chemical engineering,
mechanical engineering and environmental physics.
Organizers: Johannes P. Schloder, Universitat of Heidelberg, Germany; and
Stephen J. Wright, Argonne National Laboratory
- Feasible Point Trust-Region Methods for Equality Constrained Least Squares Problems and
Application to Parameter Estimation in Nonlinear Models with Singularities
- Hubert Schwetlick and Stefan Schleiff, Technische Universitat Dresden, Germany
- Global Optimization of Functionals Constrained by Differential Equations: Bayesian Search
on Approximants
- Prasana Venkatesh, University of Minnesota, Minneapolis
- Optimum Experimental Design for Nonlinear Dynamic Processes: Methods, Algorithms, Applications
in Robotics and Chemical Kinetics
- Klaus-Dieter Hilf, Universitat Heidelberg, Germany
- Efficient Numerical Methods for Parameter Estimation in Nonlinear 2D Transport Reaction
Processes
- Matthias Ziesse, Universitat Heidelberg, Germany
LMH, 3/15/96