Session Two
For each of the items
within the four themes, define the problem(s)
and the potential solution(s):
Theme 1. Working with other fields
Kalyan: No common language even in the EC field. We should at least
do it within the community first, e.g., putting it on the web etc.
Xin: Should stop
using the terms from biologist, that is where the
confusion arising.
Bob: Produce tools
where everyone could work together and establish the
common terminologies.
Dan: The problem
on the lack of terminologies also arises in biologist. It is
still manageable as long as we could manage the conflict.
Xin: Quite successful so far. However, we should understand how the
success of hybridization was established?
Kalyan: From engineering
or real-world point of view, the focus would be
on finding a solution instead of the convergence, etc.
Xin: There are various
ways to look at EC. Present hybridization needs to
know the domain knowledge before the hybrid methods are applied.
Hussein: Needs the sort of people (mediator) who can bridge the experts
together.
Janet: Agreed with
the point that establishment of mediators is important.
Dan: To get students
study in a second or a multi-disciplinary area could be
useful.
Theme 2. Getting industry on-board
Gary: Proactive approach and talk to them. The hardest is to establish
the first contact. Persistent and need to convince them that EC is a
solution in addition to their existing approaches, with prototype and
good demonstration.
Dan: Keep track
on the development of students who have graduated.
Organize tutorials and workshops for industry applications.
Kalyan: To collaborate
with other schools could be useful.
Bob: IPR issues,
e.g., successful industrial applications cannot be published.
Hussein: Need to
establish a prototype in order to convince the industrial
people.
Dan: To include industrial applications in the test suites.
Gary: It is often not
easy to get real-world data from industry, although
sometimes past data could be available.
Audience: Should have
more retraining and part-time degrees for people
from the industry.
Xin: Various re-training schemes have been available in UK for several
years but with limited success. One approach is to employ full-time
industrial-academic relation managers to establish the networking.
Dan: Agreed it with
similar experience encountered in the US.
Kalyan: Industry
in India often wants quick solutions for hard problems.
Dan: Different funding
for different needs is available in US, e.g., consultant
works for quick- solution problems; else, the funding could be based on
contracts.
Theme 3. Education issues
Kalyan: For EMO, what is needed is only a single solution; there is the
problem of selecting the solution from the Pareto-front. Other issues
include the convergence problem. For the schema theorem, lecturers or
students often try to avoid it due to its complexity.
Xin: Agreed and
mentioned that this is a generic problem in computer
science as their background is in logics etc., where as EC requires statistics
and calculus. One solution is to train a subset of postgraduate students, e.g.,
MSC programme in Natural Computing offered in University of
Birmingham.
Hussein: Try to be aware
of other methods for solving a problem besides
EC. Hence avoids the possibility of trying to re-inventing the wheel.
Dan: Quality of
student is low. That may be due to the low pay of teachers.
Hence we end up having to teach students everything from the basic.
Audience: The problem
of publishing papers due to the sake of getting the
tenure.
Janet: One way to ensure
the quality is through the citations.
Bob: “Vision”
of a person could be an important criteria to be judged in
assessing the applicant.
Kalyan: Some researchers are working in many diverse areas. It may be
better to focus on more specific areas in research. This of course will
depend on the interest of individuals.
Bob: To get collaboration
with others that have different domain
knowledge could help to improve the quality of works.
Kalyan: How to get more “respect” from computer science community or
others for the EC field?
Janet: The acceptance
rate for the papers in EC is high. One way is to limit
papers to more important topics.
Xin: If someone
feels that EC is easy in getting publications and funding,
one way is to challenge them to publish in EC transactions while we publish
in journals of their areas.
Theme 4. Is EC the answer?
Audience: Can establish set of benchmark problems and competitions
to evaluate the scalability of EC algorithms.
Kalyan: To organize
workshop on test problems in future EC meetings,
which can also include researchers from classical approaches.
Xin: For combinatorial
problems, there are a set of test suites for evaluating
algorithms, e.g., TSP for bin packaging, data mining (KDD and UCI
machine repository) etc.
Bob: Complex system
problems could be useful in this aspect.
Audience: Can we
establish some sort of APIs like the Java API and make
it available to others?
Kalyan: We need to tackle
problems that have not been solved. Scalability is
important: EC methods have often been applied to existing problems only.
Dan: While time complexity
in EC could be difficult, it may not be a
problem in sociology.
Bob: “Creativity”
is important in producing the framework.
Audience: Need to establish
the criteria in defining the complexity in EC.
Kalayn: This leads back
to the terminologies issue.
Xin: The first question
for applying EC to combinatorial optimization
problems is the time complexity of the algorithm. Although it is possible to
get the worst case scenario, it is difficult to get the mean of the solutions. It
is difficult to show time complexity in EC.
Dan: Theory developed is often only applicable to simple problems, but
not for complex/large scale problems. Perhaps we could borrow existing
theories from evolution.
Gary: The present techniques
of EC already borrow theories from evolution.
Dan: Right now we are
mainly getting inspiration rather than a direct use of
evolution theories.
Gary: There remains to
be plenty of theories from evolution that we can
borrow.
Xin: There is a balance
in terms of applications and theory in CEC.
Gary: Need to have EC
applications in order to convince others and to
survive in the field. Only a few researchers in biologist are convinced of EC
now.
Ali and Gary: This has been planned in future CEC conferences.
1. CEC'2002 FDEC Workshop
