|
2004 MCS Divisional Seminars & Colloquia |
|
Towards the Solution of Large-Scale Nonlinear SDP
Problems
|
|
| Abstract | Memory requirements and complexity of Hessian assembling are limiting
factors when solving large-scale linear SDPs by second-order methods. The situation becomes dramatic when trying to solve nonlinear
SDPs. We propose a first-order method that is based on the Augmented Lagrangian
algorithm whereas the linear systems are solved by the preconditioned conjugate gradient method. The new algorithm is implemented in the code
PENNON. We will comment on computational experience for linear and nonlinear
SDPs, and serve donuts to the audience. This is a joint work with Michael Stingl, from the University of Erlangen. |
| [MCS | Research | Resources | People | Collaboration | Software | Publications | Information] | |||
|
