3:15 PM-5:15 PM
Law School Room 290
Large-scale optimization problems are prevalent throughout computational science and engineering. In many cases the most time consuming aspect is the design and execution of codes to evaluate derivatives, e.g., gradient vectors and Hessian matrices. The theme of this session is how to adapt, design, and structure problems and approaches to allow for the efficient use of automatic differentiation tools in the solution large-scale optimization problems. Robust automatic differentiation software is now available. Unfortunately, "blind" use of current AD software can be grossly Inefficient with respect to both space and time. Current research efforts are aimed at modifying AD methods and software for large-scale problems, and adapting AD technology to exploit structure.
This minisymposium is concerned with recent trends in applying AD technology to large-scale optimization problems. Sparsity and structure of large application classes underlie many of the techniques being developed for efficiently using AD technology in the large-scale optimization setting. The speakers in this session will discuss the structure of large-scale optimization problems and how AD methodology is now being developed to exploit this structure (to gain efficiency and practicality)
Organizers: Thomas F. Coleman, Cornell University; and Andreas Griewank, Technical University of Dresden, Germany
AN97 Homepage | Program Updates|
Registration | Hotel and Dormitory Information | Transportation | Program-at-a-Glance | Program Overview