96. Constrained Optimization in the Presence of Noise
Invited abstract in session WB-6: Structured nonsmooth optimization -- Part II, stream Nonsmooth and nonconvex optimization.
Wednesday, 10:30-12:30Room: B100/7013
Authors (first author is the speaker)
| 1. | Figen Oztoprak
|
| Gebze Technical University |
Abstract
In this work, we consider solving nonlinear optimization problems when only noisy evaluations of the objective and constraint functions are available. In particular, we consider two variants of the line-search sequential quadratic programming (SQP) method. The first variant is designed to work with noisy equality constraints, whereas the second one is designed to work robustly when noisy inequality constraints exist. We give convergence analysis for both methods under the assumption of bounded noise. We also present numerical experiments to give insights on the practical behavior of those methods, and to highlight the issues to be solved in a practical implementation (such as the termination criteria and the computation of quasi-Newton approximations). We motivate the problem setup studied in this work with an application example from robust design optimization. Another important application of our work is in the field of derivative-free optimization, when finite differences are employed to estimate gradients.
Keywords
- Linear and nonlinear optimization
Status: accepted
Back to the list of papers