The goal of this project is to study filter methods for nonlinear optimization. These methods represent a new class of globalization techniques, that is techniques which guarantee global convergence (i.e. convergence from any starting point) of Newton's method, while impairing the natural efficiency of the underlying local algorithm as little as possible.
|Effective start/end date||1/09/02 → 31/08/07|
- nonlinear optimization
- filter methods