In any branch of engineering and science, as soon as our ability to quantitatively model phenomena reaches a practically useful level, there is a role for optimization. Numerical optimization can occur in the design and planning process by helping to select the most desirable characteristics of a design based on a quantitative model, and in the modelling process itself by helping to choose the best value of parameters needed by the model.
In this course we will focus on the computational aspects of optimization with constraints. That is, we will consider the case where we are given a well defined objective function, and certain constraints that must be satisfied. In essentially all practical situations, numerical algorithms must be used to solve these problems. We will consider some important classes of constrained optimization problems. We will learn how algorithms to solve them are constructed and to what extent these algorithms may be relied on. Both theoretical and experimental algorithmic issues will be considered. We will learn about and use some publically available optimization software. There will be regular assignments and a final paper.
The princpal prerequisite is a course in numerical computation. Familiarity with linear algebra and programming ability are important.
The main topics to be covered are:
The text for the course is " Numerical Optimization" by J. Nocedal and S. Wright
Lectures: 12:30-1:45 Tuesday and Thursday in ECCR 831
Instructor: Richard Byrd, ECOT 620, richard@cs.colorado.edu
Office hours: Monday 2-3:30, Wednesday 10-11:30
Assignments (postscript)
Assignment 1 . due January 24
Assignment 2 . due January 31
Assignment 3 . due February 7
Matlab tutorial (from Fosdick, Jessup, Schauble and Domik book)