Panel Discussion on GP thesis I wish I had written
Wednesday, 16 July, 1997, 11:30 - 12:30 pm
Genetic Programming
(GP97),
Stanford, California, USA
Panel Members
The panel is drawn from people who have recently completed PhD thesis
in genetic programming.
While we have an official title, the aim of the panel discussion is
that it should be of interest to curent and potential students aiming
to get a PhD related to genetic programming and their current or
prospective supervisors/advisors.
Walter Tackett received his Ph.D. in computer Engineering in 1994 from
University of Southern California.
His dissertation, entitled: "Recombination, Selection, and the Genetic
Construction of Computer Programs" contained among other things the first
application of GP to a "real-world" problem.
Dr. Tackett is currently the President and alpha-geek at
Neuromedia, a company
in San Francisco which makes Intelligent Agent authoring tools. While his
activities are AI related, his current work does not follow GP or any other set
technological doctrines.
Conor Ryan, University College, Cork
Shane Bruce, Nova Southeastern University, Fort Lauderdale, Florida
Summary
There were three short presentations
(by Christian Jacob, Walter Tackett, Justinian Rosca)
followed by a lively discussion sparked by questions from the
audience.
The discussion included the finicial gain in having a PhD, with views
such as "I'm not in it for the money!" to it doenst pay enough and
ways to combine GP, Java and the internet to make a fortune. Other
topics included how long it takes to write up, with answers such as
more than you expect and plan it (and stick with your plan) so you get
it done before your funding runs out. The wider topic of the
reasoning behind requiring a PhD for some teaching appointments and
the amount of teaching experience holders of PhDs have, was also
raised.
Suggestions for GP Students
- What message do you want to give to students writing
dissertations related to Genetic Programming?
WAT
-
Major potential topics
UM
WAT
WBL
-
For one or all of the topics in 1., elaborate in terms of
suggesting several specific investigations, a methodology, a problem
class, related work, etc.
UM
WAT
WBL
-
What are good GP problems for a dissertation? By ``good'' let's
mean useful for illustration and evidence that the concept you have
for extending our knowledge about GP is worthwhile.
UM
WAT
WBL
-
What overused or unillustrative problems detract from convincing
you about a GP research project?
UM
WAT
WBL
-
What is the most difficult aspect of GP of which to convince skeptical
committee members? How did you convince them? What do committee members find
intuitive about GP? Which of their intuitions mislead them?
UM
WAT
WBL
-
Are there any areas in GP research you would steer a
dissertation-student away from? Why?
UM
WAT
-
Comment on what you consider a solid methodology for
experimentation and investigations based on GP.
UM
WAT
-
How do your GP projects fit into the broad investigative
perspective of your work or research?
UM
WAT
-
What tools, advice and references did you find particularily
useful in preparing your dissertation?
UM
WAT
WBL
-
Do you have advice on computational resources, software platforms?
UM
WAT
WBL
-
Other comments?
UM
WBL
I would like to thank other holders of PhDs in GP who contributed
suggestions for the panel even though they were not able to attend,
especially Una-May for devising the above questionair and the title
for the panel discussion.
W.B.Langdon@cs.bham.ac.uk
25 July 1997