Return to Colloquia & Seminar listing
On Solving Trust-Region Subproblems
Student-Run Optimization and Variational AnalysisSpeaker: | Johannes Brust, UC Merced |
Location: | 2240 MSB |
Start time: | Wed, Mar 9 2016, 4:10PM |
The problem of minimizing a vector valued scalar objective function can be approached with a Trust-Region algorithm.
This method is based on minimizing Taylor approximations of the objective function subject to a constraint. These constrained problems are the so called Trust-Region subproblems. Our focus is on solving the subproblem when the number of unknowns is large, and the 2nd derivative matrix of the objective function is difficult to compute.