The LP-Newton Method and Recent Developments Event
- Time:
- 13:00 - 14:00
- Date:
- 22 March 2018
- Venue:
- Building 54, Room 8031 (8C)
For more information regarding this event, please email Allain Zemkoho at A.B.Zemkoho@soton.ac.uk .
Event details
The LP-Newton method was designed for systems of nonlinear equations possibly having nonisolated and degenerate solutions. Such systems arise from Karush-Kuhn-Tucker conditions for optimization problems, variational inequalities, or generalized Nash equilibrium problems. Under mild conditions, the subproblems of the LP-Newton method become linear programs. Appropriate assumptions guarantee that the LP-Newton method converges locally with quadratic rate even for cases with nonisolated and degenerate solutions. Besides some basic results, we will highlight the role of error bounds and present applications to piecewise smooth systems and generalized Nash equilibrium problems. Moreover, we will detail ideas for the globalization of the LP-Newton method.
Speaker information
Andreas Fischer ,Technical Unversity of Dresden ,: After finishing the doctorate and habilitation at TU Dresden, Dr. Fischer became associate professor at the University of Dortmund in 1998. Four years later, he got the Chair of Numerical Optimization at TU Dresden. Since 2013 he is the director of the Institute of Numerical Mathematics. His research concentrates around the design and analysis of efficient algorithms within the field of mathematical programming, especially for constrained optimization and complementarity systems including generalized Nash equilibrium problems. In addition, he is working in applications from engineering sciences with links to machine learning, discrete optimization, or structural optimization. Currently, he is PI or Co-PI of several funded research projects.