Saturday, September 23

MS30
Large-Scale Nonlinear Programming: Algorithms and Applications

10:30 AM-12:30 PM
Mt. Vernon

With the remarkable increase in computational capability, there has been a concomitant increase in the desire to apply optimization techniques to nonlinear mathematical models of phenomena arising from realistic engineering, science, and business applications. The problems under consideration in this session are characterized by having a large number of variables and/or constraints, thus the algorithms seek to exploit parallel and distributed platforms. Two speakers discuss gradient-based methods, in particular SQP methods, for solving PDE-based optimization problems. The third addresses optimization methods for problems where derivatives are unavailable, and the fourth deals with the the important problem of bounds constraints.

10:30-10:55 Parallel Software for Large-Scale Optimization

Steven J. Benson, Lois Curfman McInnes, and Jorge Moré, Argonne National Laboratory, USA

11:00-11:25 A Class of Trust-Region Methods for Simulation-Based Optimization

Juan C. Meza and Patricia D. Hough, Sandia National Laboratories, USA

11:30-11:55 Parallel Algorithms for Large-Scale PDE-Constrained Optimization

Volkan Akcelik, George Biros, Ioannis Epanomeritakis, and Omar Ghattas, Carnegie Mellon University, USA

12:00-12:25 Tightly-coupled SQP Algorithms from Loosely-Coupled Components

Paul T. Boggs, Organizer