|
2004 > |
Course Description |
Adaptive
Individuals in Evolving Populations
This course will first introduce
techniques of evolutionary computation (also
know as genetic algorithms) as models of evolving populations for use in
cognitive science. It will then explore new questions arising as these models
are combined with connectionist models of individual learning and grammatical
models of maturation (growth). Finally,
the approach will be applied to construction and control of artificial robots.
Primary text: An Introduction to Genetic Algorithms (Complex Adaptive Systems)
by Melanie Mitchell
MIT Press, 1999
ISBN: 0262631857
Day 1:
Introduction to the evolutionary approach
Required
readings:
Melanie Mitchell. An Introduction to Genetic Algorithms
(Complex Adaptive Systems), 1.9
Melanie Mitchell. An Introduction to Genetic Algorithms
(Complex Adaptive Systems), 1.1-1.6
Lab
session: Python
Day 2:
Representations & operators
Required
readings:
Melanie Mitchell. An Introduction to Genetic Algorithms
(Complex Adaptive Systems), 5.2,5.3
Melanie Mitchell. An Introduction to Genetic Algorithms
(Complex Adaptive Systems), 5.4-5.6
Melanie Mitchell. An Introduction to Genetic Algorithms
(Complex Adaptive Systems), 3.2-3.4
Lab
session: Basic Evolutionary Computation
Day 3:
Evolution & learning
Required
readings:
Melanie Mitchell. An Introduction to Genetic Algorithms
(Complex Adaptive Systems), 3.1
Melanie Mitchell. An Introduction to Genetic Algorithms
(Complex Adaptive Systems), 2.3,
Optional readings:
I. Harvey. The Puzzle of the Persistent Question
Marks: A Case Study of Genetic Drift
Lab
session: Role of CrossOver
Day 4:
Evolution & Maturation
Required
readings:
F. Cecconi, F. Menczer, R. Belew. Maturation and evolution of
imitative learning in artificial organisms
Optional readings:
P. Todd, G. Miller. Parental guidance
suggested: How parental imprinting evolves through sexual selection as an
adaptive learning mechanism
Lab
session: Evolving functions
Day 5:
Evolving robotic designs
Required
readings:
C. Mauter, R. Belew. Evolving Robot
Morphology and Control
Optional readings:
G. Hornby, J. Pollacj. Creating
High-level Components with a Generative Representation for Body-Brain Evolution
Lab session: Evolving
functions II
Assessment:
Lab sessions
will explore simple implementations of evolutionary algorithms in Python. Students wishing to take the course for
credit are expected to participate in these lab sessions.
Richard K. Belew is a professor
of Cognitive Science at The University
of California, San Diego. His research
focuses on adaptive knowledge representation,
particularly the integration of induced knowledge with traditional
scientific publication and data. He is
author of Finding Out About (Cambridge Univ. Press, 2000) and co-editor (with
Melanie Mitchell) of Adaptive Individuals in Populations (Addison-Wesley,
1996). He received his PhD from the
Unviersity of Michigan..