|
An Efficient Sampling Approach to Multi-objective
Optimization under Uncertainty
Urmila Diewekar
University of Illinois at Chicago
Hosted by Mihai Anitescu
3:00 PM, January 14, 2004
Building 203, Room C230
|
| Abstract |
Robust decision making under uncertainty is of fundamental importance in
numerous disciplines and application areas. For many practical issues, decision making involves multiple, often conflicting, goals and poses a
challenging and complex optimization problem. The problem of decision making under uncertainty is posed as a stochastic optimization problem,
which fundamentally involves constrained optimization of one or more probabilistic output functions constructed from multiple deterministic
simulations for input parameter sets obtained by sampling uncertain input parameter distributions. Although valuable, the computational burden of
this approach can be extreme and depends on the sample size used for characterizing the parametric uncertainties. The computational tedium
presents a critical impediment to a widespread use of the stochastic analysis and
optimization approach to robust decision making.
The overall objective of this talk is to present innovative computational
strategies for significantly advancing the state-of-the-art in the area of optimal decision making under uncertainty by (a) improving algorithms
for sampling over uncertain variables, (b) developing a novel and generic approach to quantitatively characterize the accuracy of various sampling
techniques using principles of fractal geometry, and (c) incorporating the knowledge of the accuracy of sampling in the stochastic optimization
process thereby automating and improving the optimization algorithms.
Further, this talk presents recent innovations in the techniques underlying multi-objective optimization, in the characterization of
uncertainties, and in the ability to develop and apply these methods outside of traditional application domains greatly enhances their utility
and promise.
|
