Teaching Material EP'95

1
Christoph F. Eick
Transparencies Course COSC 6367 (Spring
2001) Evolutionary Programming
James Rosenquists Evolutionary Balance (1977)
2
Course Description Evolutionary Programming

• Evolutionary Programming Cr. 3 (3-0).
  Prerequisites graduate standing, programming
  experience, MATH 3336 or consent with instructor.
  Theory and application of evolutionary
  programming and other related areas in
  evolutionary and natural computation centering on
  genetic algorithms and programming, evolution
  stategies, artificial life, and other models that
  rely on evolutionary principles. Students will
  perform course projects that apply the discussed
  techniques to numerical optimization problems, to
  machine learning, and to the simulation of
  biological and cultural systems.

3
EP/GA/EC/EA

• Evolutionary programming, for short EP, (also
  sometimes called genetic algorithms(GA) and
  Evolutionary Computing(EC)) steals the idea of
  evolution from biology sets of solutions are
  evolved by applying genetic operators such as
  crossover (a new solution is created by mating
  two parent solutions), and mutation (random
  change in a solution).
• EP is getting more and more popular more
  conferences publish EP papers
• EP technology is successfully applied in natural
  sciences and in engineering to complex problems,
  such as
• prediction of chemical structures in 2D and 3D
• optimization in chemical engineering
• scheduling problems
• simulation of biological systems
• EP techniques are also intensively used in
  machine learning research. Popular applications
  include
• knowledge discovery in databases
• learning class desciptions from sets of examples
• genetic programming (learning programs through
  evolution)
• learning strategies in multi-agent environments

4
The Puzzling World of Evolutionary Computation
5
Goals of COSC 6367

• Prerequisites the class is basically
  self-contained the only skill you are assumed to
  have are basic programming skills (undergraduate
  data structures level is enough).
• Topics Covered
• introduction to evolutionary programming
• review of "classical" machine learning,
  optimization, and heuristic search techniques
  learn how to design and implement evolutionary
  programming systems that focus on those tasks.
• practical applications of evolutionary
  programming to optimization and machine learning
  by doing projects.
• introduction to artificial life
• learn how to conduct a small research project
  learn how to present empirical results learning
  how to summarize and present project results
  (this might be quite important for your current
  and/or future job and for your dissertation/Master
  's thesis)
• Teaching Style project-oriented, learning by
  doing teaching style

6
Special Remarks COSC 6367for Spring 2001

• The way the class will be taught will depend on
  the class size and available teaching resources.
  Therefore, I will select course projects and
  other class activities by January 31, 2001.
• Teaching material will be moved from an offline
  status to an online status --- but this is work
  and takes time.
• In addition to the textbook, many other sources
  will be used for teaching the class.
• Some new textbooks came out recently (some
  material in these book might be used to replace
  teaching material that was used in the last
  teaching of the class).
• Due to the last 4 reasons a detailed teaching
  plan will not be available before February 16,
  2001.
• There will be an Artificial Life group activity
  and possibly a second group activity.

7
Topics of Evolutionary Programming

• Introduction to Genetic Algorithms (GA) and
  Evolutionary Programming (EP). Theory of GAs.
  Genetic Operators. Application of GA/EP
  technology to
• Numerical and Symbolic Optimization Problems
• Heuristic Search
• Machine Learning and Knowledge Discovery
• Brief Introduction to the Field of Artificial
  Life centering on
• Cellular Automata and Emergent Behavior
• Simulation of Biological Systems
• Computer Viruses and other Computer Life Forms
• Composition of the class
• Evolutionary Programming/Genetic Algorithms
  70
• Artificial Life
  15
• Other (Heuristic Search, Machine Learning, Data
  Mining) 25

8
Elements of Evolutionary Programming

• 21-23 lectures
• 3-4 videos centering on Artificial Life and
  Genetic Programming
• 2 quizzes and one main exam
• 2 EP-Project (1 centering on optimization and one
  on machine learning)
• 1 Artificial Life Project
• Text Book Zbigniew Michalewicz Genetic
  Algorithms Data Structures Evolution
  Programs Springer-Verlag third, extended
  edition.
• High Qualify Teaching Material will be available
  (PowerPoint 4.0 slides, a good textbook, handouts
  taken from other books) that facilitates learning
  the topics that were covered in class.
• Projects will require programming (any
  programming language can be used for the
  projects), students should have programming
  experience.

9
Reading Material Covered by the Class

• Zbigniew Michalewicz Genetic Algorithms Data
  Structures Evolution Programs, Third Edition,
  Springer 1996. The course covers chapters
  1,2,3,4,5,6,8,9,10 more or less completely, and
  the chapters 7 and 11 partially.
• David Goldberg Genetic Algorithms in Search,
  Optimization, and Machine Learning, Addison
  Wesley, 1989. The course covers chapters 5
  Advanced Operators and Techniques in Genetic
  Search and 6 Introduction to Genetic-Based
  Machine Learning of the book.
• Melanie Mitchell Introduction to Genetic
  Algorithms, MIT Press, fourthcoming, to appear
  Fall 1995. The course covers section 1.9.2 Host
  and Parasites Using GAs to Evolve Sorting
  Networks, section 2.2.2 Evolving Cellular
  Automata moreover, it covers chapter 4
  Theoretical Foundations of Genetic Algorithms
  partially.
• John Koza Genetic Programming II --- Automatic
  Discovery of Reusable Programs, MIT Press, 1994.
  The course covers chapter 2 completely, and
  chapters 3 and 4 partially.
• More material to be added!!

10
Important GA/EP Conferences

• ICGA --- International Conference on Genetic
  Algorithms is held every odd-numbered year.
• EP --- Evolutionary Programming Conference is
  held every year.
• IJCAI --- International Joint Conference on
  Artificial Intelligence the main AI conference
  (usually has a special GA/EP session).
• ISMIS --- International Symposium on
  Methodologies for Intelligent Systems is usually
  held twice in three years.
• TAI --- Tool with AI usually publishes some
  EP/GA papers.