|
Description: |
Interior-point methods are a very promising option for the solution of large-scale nonconvex optimization problems. The development of these methods is already well-advanced for the treatment of linear, quadratic and general convex problems. Recently, some algorithms have been proposed for the nonconvex case, but several open questions still remain.
In this talk we will present a brief overview of the main approaches that have been considered for this nonconvex case. We will also describe the main ideas behind interior-point methods for convex and nonconvex problems. For the specific case of line search methods, several issues will be analyzed in greater detail, such as the barrier transformation of the problem, the choice of merit function, the use of negative curvature to improve the behavior of these methods, and the updating of the parameters.
Finally, we will describe some aspects of a specific implementation of an interior-point method using negative curvature. Particular attention will be paid to the conditions under which the negative curvature information is used, the way in which the descent and negative curvature directions are combined, and to the updating of the barrier parameter. The convergence properties of the method will be commented, and computational results on a set of test problems will be presented. Area(s):
|
Date: |
|
Start Time: |
15:00 |
Speaker: |
Francisco Javier Prieto (Dept. of Statistics and Econometrics, Univ. Carlos III, Madrid, Spain)
|
Place: |
Room 5.7
|
Research Groups: |
-Numerical Analysis and Optimization
|
See more:
|
<Main>
|
|