mathematics
and
computer
science

LANS

LANS Publications

"The Return of the Active Set Method"

S. Leyffer

Preprint ANL/MCS-P1277-0805

Preprint Version: [pdf]

For solving nonlinear optimization problems, two competing iterative approaches are available: active set methods and interior-point methods. Current implementations of interior methods often outperform active set methods in terms of speed. On the other hand, active set methods are more robust and better suited for warm starts, which are important for solving integer optimization problems. Consequently, we have recently become interested in new active set approaches, which are reviewed in this note.