213. A sharp augmented Lagrangian-based method for solving constrained DC optimization
Invited abstract in session WB-6: Structured nonsmooth optimization -- Part II, stream Nonsmooth and nonconvex optimization.
Wednesday, 10:30-12:30Room: B100/7013
Authors (first author is the speaker)
| 1. | Sona Taheri
|
| RMIT University |
Abstract
In this talk, we will focus on constrained difference of convex (DC) optimization problems. Most existing methods for solving such problems do not guarantee finding feasible stationary points if the starting point is not feasible. We propose combining a method based on the sharp augmented Lagrangian with local techniques to solve constrained DC optimization problems. It is well-known that a subgradient of the dual function, formulated using sharp Lagrangians, increases the dual function, making these methods effective at quickly finding a feasible point. From there, a local method can be applied to compute a feasible stationary point. We develop a two-step method incorporating these approaches and study its convergence. Additionally, we compare its performance with a number of constrained nonsmooth optimization solvers using standard test problems.
Keywords
- Non-smooth optimization
Status: accepted
Back to the list of papers