Reminder: A final project "should be a significant effort in developing some idea
from the class further" - either a paper, a program, a physical system,...
(see Announcements and News)
In all, you will be handing in an initial proposal and one written report per group. Each group
will also give an oral presentation towards the end of the term.
The proposals are due by 5PM the Friday before spring break (March 21).
A few suggestions:
Pick a pair of related papers and compare and contrast the
approaches. (Generally not a winning option- it may sound simple,
but it will take a lot of work to make your paper contain evidence
that you have some new perspective on the topics and a deep
understanding of the details.)
Use one of the systems mentioned in lecture as your starting point:
Tierra, Conway's Life... Build from there.
Greatly extend your work on either Braitenberg vehicles or the ants
trail - really extend, don't just provide a nice graphical interface.
Find a topic mentioned in lecture and write a simulation for it
(e.g. Pigeon-homing).
Implement a system described in one of the papers listed at the end
of the course notes (we can provide you with specific references if you know
what topic you're interested in), run some tests and critique the system
constructively.
Una-May O'Reilly (she taught the lecture on Unconventional
Computation) has a
page
with interesting suggestions that she compiled in 1999. They are
still worthy ideas. But beware that some of the links may be stale.
Una-May will lecture on Evolutionary Computation in late April. There
are a great deal more software implementations of genetic programming,
genetic algorithms and other EC algorithms around now. Use Google or
any other search engine with appropriate terms and you will find
them. If you decide to use code from a public source, your project
should go further than running it and showing results. Be sure to have
a sound hypothesis and investigational goals and include sufficient
in-depth analysis.
Other ideas can be found in interesting books and proceedings such as
"Swarm Intelligence: From ComplexSwarm Intelligence : From Natural to Artificial Systems" by Eric Bonabeau, Marco Dorigo, Guy Theraulaz.
GECCO (Annual Conference on Genetic and Evolutionary Computation held in 2000, 2001).
Journal of Genetic Programming and Evolvable Hardware (Kluwer)
Don't be shy of trolling interesting research group pages on the web too. The research projects listed are of a larger scope than 6.836 projects but they are inspiring. Here's some examples:
Build a physical robot or vehicle that proves something about the
importance of embodiement rather than simulation.
Think up an application for an artificial creature; design it on
paper, specify its behavior, explain why it'd be a good application of
the technology.
Do something that's never been done before.
These suggestions are neither exhaustive nor exclusive. You can talk to
any of us about your ideas for a project, and get feedback
on its shape and scope.
