| |
|
Description: |
In this talk we show how the filter technique of Fletcher and Leyffer (1997) can be used to globalize a primal-dual interior-point Newton method for nonlinear programming. Filter methods apply the concept of dominance from bi-criteria optimization to handle the competing aims of minimizing the objective function and an infeasibility measure at the same time. Thus, the filter approach avoids the use of merit functions and the update of penalty parameters.
We present a new primal-dual interior-point filter method and prove global convergence to first order critical points. Our work extends the recent convergence result of Fletcher, Gould, Leyffer, and Toint (1999) for SQP-filter methods to a primal-dual interior-point framework.
The new algorithm decomposes the primal-dual step obtained from the perturbed first-order necessary optimality conditions into a normal and a tangential step, whose sizes are controlled by a trust-region type parameter. Each entry in the filter is a pair of coordinates: one measures feasibility and centrality, and is associated with the normal step; the other measures complementarity and duality, and is related to the tangential step.
This is joint work with Stefan Ulbrich (Technische Universitaet Muenchen) and Luis N. Vicente (University of Coimbra).
Area(s):
|
|
Date: |
|
| Start Time: |
15:00 |
|
Speaker: |
Michael Ulbrich (Technische UniversitÀt MÌnchen, Germany)
|
|
Place: |
Room 5.7
|
| Research Groups: |
-Numerical Analysis and Optimization
|
|
See more:
|
<Main>
|
|
