Download CEE 5xx or CS 5xx MODERN HEURISTIC SEARCH METHODS (3

COM S 572 / CEE 509 Heuristic Methods for Optimization1
MWF 11:15-12:05, Spring 2003. Room 206 Hollister (room change)
Instructors: Professor C. Shoemaker <CAS12> and Professor B. Selman
Description 3 or 4 credits. Prerequisites: graduate standing or UG background in computing
(e.g. one of the following: COM S/ENGRD 211, 321 or CEE/ENGRD 241) or permission of
instructors.
Heuristic optimization algorithms are artificial intelligence search methods that can be used to
find the optimal decisions for designing or managing a wide range of complex systems. This
course describes a variety of heuristic search methods including simulated annealing, tabu search,
genetic algorithms, derandomized evolution strategy, and random walk. Algorithms will be used
to find values of discrete and/or continuous variables that optimize system performance or
improve system reliability. Students can select application projects from a range of application
areas. The advantages and disadvantages of heuristic search methods for both serial and parallel
computation are discussed in comparison to other optimization algorithms
Topics Covered
(4 lectures) Introduction to Search Methods and Computational Complexity
(3 lectures) Tabu Search
(4 lectures) Simulated Annealing
(3 lectures) Genetic algorithms
(2 lectures) Genetic Programming
(2 lectures) Random Walk
(3 lectures) Derandomized Evolution Strategy
(1 lecture) Applications of Combinational Heuristics to Real-valued Problems
(8 lectures) Applications of Heuristic Optimization in a Range of Areas (including
guest lectures)
(3 lectures) Evaluation of the Relative Performance of Alternative Heuristic Methods
(6 lectures) Theoretical basis for heuristic search methods and significance of results
(2 lectures) The Advantages and Disadvantages of Heuristic Search Methods for Both Serial and
Parallel Computation in Comparison to Other Optimization Algorithms.
(1 lectures) Response Surface Methods to Enhance Heuristics
(1 lecture) Heuristics for Multiobjective Optimization
Projects in Heuristics Course (students select one topic for team project, and students hear the
lectures on all the applications)
1.
2.
3.
4.
5.
1
Job Shop Scheduling (guest: Prof. Roundy, Operations Research & Industrial Engineering)
Satisfiability in Artificial Intelligence (Prof. Selman, Computer Science)
Cellular Networks (guest: Prof. Wicker, Electrical Engineering)
Protein Folding (guest: Prof. Shalloway, Biochemistry)
Time-varying optimization of Systems of Partial Differential Equations (Prof. Shoemaker,
Civil and Environmental Engineering)
(continued on next page)
This course was not offered in academic year 2001-2002.
Depending on student interest, projects may be added in the other areas including genetic
programming and/or machine learning. Ph.D. students can also propose independent projects.
Software: Students are expected to be familiar with Matlab. For the projects, pre-programmed
modules will be given to students that represent the response of the systems to changes suggested
by the heuristic optimization search. As a result, programming will not be a major effort for CS
and engineering students. The programs students write are used to explore the efficiency of
existing algorithms and of the students’ creative modifications of these algorithms.
Grading: The course grade will depend upon homework assignments, one prelim, a final, and the
project.
Class Compostion: This class draws students from Computer Science, many Engineering
departments, and a few departments outside the Engineering College. The largest enrollments
are from Computer Science, Operations Research and Electrical and Computer Engineering.
April 29, 2017