SIAM Short Course on Numerical Optimization
Sunday, May 19, 2002
Westin Harbour Castel Hotel, Toronto, Canada
Organizer
Robert J. Vanderbei, Princeton University
Description
In recent years there has been a growing interest in efficient algorithms for nonlinear programming (NLP). This interest has led to the development of several computer codes, which have proved very successful in solving broader classes of problems than ever before. Success in solving difficult NLPs depends both on the choice of algorithm and the manner in which the problem is formulated. In this short course, we will survey these recent developments. Topics will include various implemented algorithms, good and bad modeling techniques, and the growing importance of modeling languages. Explicit instruction in modeling various classes of problems will be provided. Comparative results for several algorithms will be presented with an analysis of the strengths and weaknesses of each.
Instructors
- Robert J. Vanderbei is a Professor in the Department of Operations Research and Financial Engineering at Princeton University. He is the author of the book Linear Programming: Foundations and Extensions .
- David F. Shanno is a Professor in the Rutgers Center for Operations Research at Rutgers University. His interests are in algorithms for continuous optimization. He is a co-developer of an interior-point code for linear programming.
- Hande Y. Benson is a recent PhD from the Department of Operations Research and Financial Engineering at Princeton University.
Who Should Attend?
- Staff of academic, government, and industrial institutions interested in learning how optimization technology can help them solve practical problems
- Researchers in biology, chemistry, physics, and economics that need to know optimization techniques for their work
Recommended Background
- computational linear algebra,
- calculus for functions of several variables (Jacobians, gradients, Hessians),
- notation and techniques used in optimization
Detailed Outline
- Algorithms for NLP
- Sequential Quadratic Programming Methods
- Reduced Gradient Methods
- Interior-Point Methods
- Algorithmic Issues
- Convex vs. Nonconvex Problems
- Line Search vs. Trust Regions
- Merit Functions vs. Filters
- Infeasibility Detection
- More Algorithm Issues
- Starting Points
- Honoring Bounds
- Jamming
- Modeling Front End---AMPL
- A Brief Language Tutorial
- Automatic Differentiation
- Hooking YOUR Solver to AMPL
- Modeling Matters
- AMPL vs. Subroutine
- 1010
- Differentiability
- Convex vs. Nonconvex
- Compact vs. Expanded Representation
- Beauty Contest---Algorithms will be compared on multiple criteria:
- Poise
- Charm
- Talent
- Swimsuit Competition
- Second Order Cone Programming
- Models
- Algorithms
- Semidefinite Programming
- Models
- Algorithms
Program
Morning
8:00 - 8:30 Registration and Welcoming Remarks
8:30 - 9:15 1. Algorithms for NLP
9:15 - 10:00 2. Algorithmic Issues
Coffee
10:30 - 11:15 3. More Algorithm Issues
11:15 - 12:00 4. Modeling Front End---AMPL
Afternoon
12:00 - 1:15 Lunch
1:15 - 2:00 5. Modeling Matters
2:00 - 2:45 6. Beauty Contest
2:45 - 3:15 Coffee
3:15 - 4:00 7. Second Order Cone Programming
4:00 - 4:45 8. Semidefinite Programming
Adjourn
Registration
Seats are limited. We urge attendees to register in advance. To register, please complete the online Preregistration Form (available shortly) and submit to SIAM. Registration fee for the course includes course notes, coffee breaks, and lunch on Sunday, May 19.
Location
This short course will be in Toronto. The coffee breaks will be in Lobby; lunch will be somewhere.
Designed by Donaghy's Web Consulting
Created: 7/27/01; Last Updated 8/01/01, DAR