Theme 1: Working with
other fields?
1.1 Terminologies
Both the panelist and
audience expressed that there may be a need to
bridge the gap in terminologies so as to make people talk about the same
thing in the same way.
David suggested using
Ontology to create the necessary standards for EC
terminologies. However, Dan cautions that even though Gene Ontology
may be around for some time, but not everyone will actually participates in it.
Russ indicated that terminology
has been a problem in many areas such as
in Neural Network (NN). He added that the IEEE NN council previously
established a standard for NN. However after merely one year the
terminology that was created is no longer followed as new terminologies
kept surfacing and there is a lack of bodies to monitor it closely.
Subsequently, Ali requested
the panelists and audience to cast a vote on the
need for a glossary on the EC field. A member of the audience expressed
that too many people have their own standards and terminologies and
hence it may not be necessary to have yet another entry. To this issue, Russ
and Xin also agree that it may not be necessary to create another
terminology as long as authors properly detail the EC terms according to the
context used in their paper.
1.2 EC Hybridization
It is recognized among
most members of the panel and audience that EC
hybridization has been very successful in many real world applications and
there is certainly no problem with it.
Dan felt that while there
is no problem with hybridization, the problem
however may lies in convincing people to do EC hybridization with other
established methods, as many people are unaware of its success.
Gary also express that
it might be easier to demonstrate EC hybridization to
customers as they may be more familiar with existing technologies that have
been used in the EC hybridization. At the same time, it allows us to show
how the use of EC hybridization can perform better than traditional
methods to the customers.
In addition, a member
of the audience indicated that he would like to see
research on EC hybridization with established statistical methods.
1.3 Domain experts
+ EC experts
Here, the need to develop
more multi-disciplinary and cross- discipline
human experts in specific domains and EC has been discussed upon. It is
felt that there exist communication issues between people from different
disciplines.
Paul stressed that it
would be a big challenge to make student more multi-
disciplinary as such as course program would be difficult to create. Rather, it
is often through on job training that one becoming more multi-disciplinary.
For example in Test Advantage, Extreme programming where pairs of
experts were made to work on the same application was used to allow them
to interact and learn from each other in different disciplines.
Dan added that sabbatical
or attachments for students at labs of different
discipline would also be helpful and such a system has been a success in his
university. Russ also felt that projects involving inter-disciplinary areas may
also be useful and we should encourage more of these in the university. To
this issue, the audience expresses the importance of horizontal and vertical
integration between the experts, in particularly, technical and managerial
integrations.
In addition, it is also
indicated by Arthur that the issue may be a challenge
but not really a big problem to the industry. Ken also pointed out that it is
not feasible for one to be expert of every application areas rather we should
focus more on establishing the EC fundamentals in the students like what
we have been doing in the past.
Theme 2: Getting industry
on board
Paul Tabor of Test Advantage
started the workshop by giving a speech on
the Evolutionary Computation (EC) Opportunities and Challenges in Semi-
conductor Manufacturing. In particular, he outlined and discussed the
driving forces, challenges, applications, commercial solutions and market
overview of EC in Semi-conductor Manufacturing.
Subsequently, Arthur Kordon
of Dow Chemical Company shared his
experiences and vision on the future directions of EC with an industrial
perspective. In particular, he discussed on the economic benefits, expected
features, industry needs, and key application areas of EC in the chemical
industry.
2.1 Defining EC problems
Gary expressed that the
industry may not realize they have a problem at all,
and may not realize that EC can present a useful solution, or may not even
aware of EC at all.
To this issue, Paul indicated
that a lot of brainstorming is often conducted
before they could identify different types of suitable applications for their
clients, which usually take up to 6 -- 18 months before marketing.
Arthur added that the
diversity of most real world problems are often very
wide and cannot be addressed using a single approach such as EC. Instead
one should combine EC with other techniques like Neural Networks
creatively to effectively address issues in the real world environment.
Dan suggested having researchers
conduct their sabbatical at the industry to
better interact with them and learn their problems in order to be able to
better identify and solve their problems. To this suggestion, Russ and David
caution the issues of Intellectual Property rights between companies and the
universities and identify this would be the core problem.
In contrast, Arthur and
Paul expressed the importance of Intellectual
Property rights to p their companies that is crucial to help them maintain
their competitiveness.
A member of audience wishes
to know how important Evolutionary Multi-
Objective Optimization (EMO) is to the industry. Arthur replied that EMO
is top of the priority and more concentrations should be placed on it as it is
regarded as the technology capable of delivering parsimonious solution to
the industry. In addition, visualization was also suggested by a member of
the audience as having a key role to play the link between EC and the
industry.
2.2 Having convincing
applications?
It has been felt that
there is a limited number of convincing applications
published in the CEC conference.
The general feeling shared
among the panelists and the audience is that this
may be due to the fact that most convincing applications of EC are usually
published in application specific conferences or journals as they are often
viewed as more contributing in that sense.
To increase the number
of convincing applications in CEC, Dan suggested
that the industry could help by proposing more special sessions. However,
the concern of the industry is that their papers may not be accepted. Xin
added that the IEEE Transactions of EC do accept application papers of
EC but those that do not contribute directly to EC are usually rejected.
Arthur expressed that one solution to this would be to define different
criteria on the qualities of a good application paper.
Later, Xin further suggested
learning from IJCNN by publishing and
circulating a list of examples on successful EC industry applications. Further
it may be necessary to have people from the industry to provide the index
for success applications.
Theme 3: Education
Issues
3.1. Proper background:
lack of mathematics for biology/computer
science students
In this sub-theme, the
background of students in the area of EC is
discussed.
Xin described that back
in Birmingham, their master program on EC has
been quite a success. The EC course is packed with lot of stuffs on statistics
and has graduates from all disciplines, including biology and chemistry.
A member of the audience
however felt that rather than teaching the
students all specific topics in classroom, we should educate them to take
learning as a lifelong task. Hence we should perhaps educate students in
teaching themselves.
David suggested educating
the students based on the needs of the industry.
However, Paul thinks that from the industry perspective, student should be
allowed to choose their field of specialization as they deem fit. Arthur added
that the concentration should be on students abilities to do research in a
common sense way and be proficient in their communication ability as well
as capable of using simple analogy to express concepts to the customers.
Ali asked the panelist
and audience if they thought there are appropriate
undergraduate textbooks on EC?
Dan replied that there
is a lack a good book on EC for undergraduate.
Often educators have to write their own lecture notes.
3.2 Career development
and convincing research
Arthur expressed that
the industry would prefer graduates with skill set in
many areas rather than just EC. The audience agrees with Arthur and
indicated that even thought one may not continue with EC as their
immediate career but it may certainly prove to be useful at a later stage in life
since the fundamental of EC has been drilled into them through the EC
courses.
3.3 Interdisciplinary
area for students and researchers
Generally, there has been
a general consensus on the interdisciplinary area of
EC for students and researchers. Dan pointed out that there have been a lot
of EC groups in Business colleges. Further the audience also felt that more
institutions would conduct EC courses in the near future.
Theme 4: Is EC the
answer?
4.1 Is scalability
an issue?
Arthur and David explained
that scalability is important and EC has been
found to be successful and scales well with increasing problem complexity.
4.2 Balance of application
and theory
The audience felt that
theories and error bounds are generally useful for
publications and the writing of books. On the other hand, empirical studies
are generally sufficient for engineers in real world applications. Nevertheless,
it would be to useful and necessary to employ analogies to show whether or
not an EC method works well and when the proposed method would work
extremely well or even fails.
Xin pointed out that it
is often necessary to put oneself in the view of the
reader as this will help decide on the balance between applications and
theories which helps makes a high quality paper.
Ali later posed the call
for possible solutions to increase inputs from the
industry.
Bob pointed out that a
possible solution may be to host more industry
sponsored competitions.
Both Arthur and Paul expressed
that the interests in CEC lie in commercial
interest after stumbling onto EC through their researchers and great
benefits have been found since. They felt that the reason why companies
are not involved in EC is mainly due to the lack of convergence theory in
EC. Gary on the other hand pointed out that industry would start rushing
in once they begin to understand the power of EC.
4.3 Industry input
in EC meetings
The audience felt that
how much industry would be willing to contribute to
EC meetings such as CEC depends on what they could benefit out of the
CEC. Hence, there may be a need to market EC meetings to make it more
explicit to the industries. Open source code is also regarded as a possible
way to encourage more industry inputs.
Epilogue
In this section, a vote
on which topics that will remains to investigate and
discussed in the future workshop has been made?
Theme 1: Working with
other fields?
1.1 Terminologies:
Not necessary to work on it as it is not a problem, GO!
1.2 Hybridization:
A lot can still be done here. Stay.
1.3 Domain experts
+ EC experts: GO!
Theme 2: Getting industry
on board
2.1 Defining EC problems,
Stay
2.2 Having convincing
app: Stay
2.3 Fast answers to meaningful
problems: Stay
Theme 3: Education
3.1 Proper Background:
Not within our control, Stay
3.2 Career
development and convincing research: GO!
3.3 Interdisciplinary
area for students and researchers: Stay
Theme 4: Is EC the answer?
4.1 Is scalability an
issue? : No decision was made.
4.2 Balance of application
and theory: No decision was made.
4.3 Industry input in
EC meetings: No decision was made.
David Fogel pointed our
that we are in fact moving very well in the EC field
and have been very successful in fact. EC, being a small group to a bigger
IEEE CI society, which is bigger than many other societies, we are well
recognized in IEEE. Our impact factor in 2001 was 1.7.
Ali explained that the
FDEC workshop does not suggest that EC is not well
but rather we are looking for suggestions and ideas that may help advance
EC further. For example, Ali stressed that part of the results from this
workshop would be transform into a white paper which will be updated
yearly, in CIS newsletter. The conclusion drawn here would also help make
recommendations to the technical committee workgroup and decide what
the future direction of EC is.
Lastly, members of audience
felt that the FDEC workshop has been an
important component to identify the future of EC. In particular, suggestions
were made to highlight the changes in the EC field over the years and
identify the niches of present EC and how to move it forward. The lists of
possible topics may be looked upon:
1. CEC'2002 FDEC Workshop
