The goal of this project is to perform an
experiment to test a hypothesis relating to evolutionary computation
or to write an evolutionary algorithm for a specific (challenging) application.
The exact hypothesis or application is up to you
(but see subproject 4a). If you plan
to test a hypothesis make sure that
it is clear, well defined, and reasonable to answer experimentally.
Ideally you can claim "if my hypothesis is correct and if I perform this
experiment then the result will be X and if the result is not X then
my hypothesis is incorrect".
The simpler and more specific your hypothesis is, the easier it will
be to do this project, and you should understand the reasoning behind
your hyptohesis. For example, if your hypothesis is "using random
trails in the artificial ant/Sante Fe trail problem will improve
results" you're going to have a hard time - what does improve mean? Why
should they improve?
On the other hand if your hypothesis is "using
random trails in the artificial ant/Santa Fe trail problem
will on average produce individuals whose
fitness is higher, although they may take longer to evolve,
because the random trails will force GP to
evolve a more general and hence more successful soluton" it's going to
be much easier to complete the project.
If you plan to tackle a specific application you should have a specific reason
why the application is interesting and challenging.
Project Requirements:
- A clearly defined hypothesis or applcation.
- A clearly defined experiment to test
the hypothesis or a clearly defined algorithm for the application.
- The code to run the experiment. I strongly encourage you to figure
out a way to reuse code from previous experiments.
Project Write-up: You must write a short paper describing the
results of your project. The paper should be formatted using
the
ACM SIG Proceedings Templates. This is a standard format for many research
conferences. Note that there is a Word and a Latex template, you may use
either one.
The final paper should include the following sections:
- Abstract - a short (~200 words) summary of what you did
and
what the results were.
- Introduction - including: the hypothesis or application,
what evidence you
have for and against the hypothesis or previous work on the applicaiton, include at least
3 published sources.
- Experiment description - including:
- The test problem(s) used in the experiments.
- A description of the evolutionary algorithm used in the experiments.
- How fitness was measured.
- What parameters were used.
- If you are testing a hypothesis you should include a
clear explanation of how the results would support, or refute,
the hypothesis. You should be able to say, before running any
experiments, 'if I get these results it means the hypothesis is
confirmed (or at least supported) and if I get these results the
hypothesis is refuted'.
- Results:
- There should be at least 5-10 trials per experiment.
- Include graphs and/or tables to make it easy to understand the results.
- Make sure that the graphs and tables are clearly labeled.
- Explain how the results support or refute your
hypothesis or how well you algorithm did on the application problem.
- Conclusions
- References