Presented at the 1997 CRPC Annual Meeting
Numerical optimization can have enormous computational requirements, especially in engineering and scientific applications where the optimization problem encompasses a complex simulation that describes the behavior of the system being optimized. We will discuss how various optimization algorithms work and how they can be effectively parallelized.
This tutorial will focus on general linear and nonlinear optimization problems with continuous variables. We will not cover combinatorial optimization, boolean programming, or integer programming in any depth.
Topics will include
- Specification of optimization problems.
- Various kinds of optimization algorithms and how they work.
- Guidelines for matching an optimization algorithm to a problem.
- The structure of various optimization algorithms as they are implemented.
- What parallelizes, and what doesn't.
- Scalability of parallel optimization approaches.
- Software resources.
This tutorial is intended for people who are interested in applying parallel optimization methods in engineering and scientific problems. The course presumes a basic knowledge of elementary calculus and linear algebra; familiarity with at least one optimization technique is also useful, but not necessary.