SFI/CSSS'07 Evolutionary Algorithms Lecture Topics

Una May O'Reilly asked:
Hi member of CSSS class of 2007, I'm working on lecture material for Week 3 when I'll be up on Tuesday and Thursday. I'm going to be lecturing on Evolutionary Algorithms. If you have time, can you drop me an email wrt to questions below, please:
  • What is your knowledge/experience level with respect to genetic algorithms, genetic programming, evolutionary strategies, evolutionary
    programming, particle swarm optimization, BOA, hBOA, grammatical evolution or any other Evolutionary Algorithm? Responses might be one of "expert, knowledgeble, ignorant" and one of "develop, use, don't use" for each of these.
  • How you have used or anticipate using evolutionary computation in your field of research?
  • Are there any topics or concepts in evolutionary algorithms that are burning a hole in your head that you'd like me to cover?
I easily fall into the nearly ignorant category with almost no knowledge of genetic algorithms, etc.  I've seen a couple of articles about them in the general tech/science press (e.g. Technology Review) and have scanned some websites about the topic but that is about it.
 
I don't generally do research myself but hope to find worthy areas of pursuit at the intersection of the needs of public interest, the goals of our government sponsors, and the interests/abilities of the staff at our research center.  I don't think my reserch center has done any research drawing on this topic but it appears a good candidate for pursuit so am very intereted in the topic from this perspective.
 
Some discussion or coverage of these topics might be valuable to for us: 
 - utility in finding approximate solutions to real-world problems as opposed to the "best" solution
 - case studies and references such as the GE/RPI jet engine example or Andreaou et al on predicting foreign exchange rates,
 - applications to types of problems such as scheduling a broad and long enterprise initiative or inspections for hazardous materials
 - an explanation of how the principle of "fitness" is applied and suggestions on how to describe superior fitness functions
 - comparison to common constraint satisfaction techniques (ex: http://aima.cs.berkeley.edu/newchap05.pdf)
 - dealing with problems where it would be difficult, even impossible, to specify other than an arbitrary initial population
 - applications (if any) where the population is very dynamic
 - recommendations on the best general and special topic references for this area of study would be very valuable
 - popular puzzle metaphors are valuable - e.g. http://en.wikipedia.org/wiki/Hare_and_Tortoise or http://en.wikipedia.org/wiki/Queens_problem
 - comparison and contrast to other alteratives would be great (neural networks, annealing, etc.)
 - principles for preventing premature or sub-optimal outcomes
 - applications that do not [sic] take advantage of, or require, spectacular improvements in CPU processing power
 
R/
 
 -  Brian Lawler
 

No comments: