Discussing the idea of a special issue on CoEC
R. Paul Wiegand
> One concrete suggestion to make the foundations of
> coev. algorithms more widely considered (in particular,
> better known to IEEE people, if there is such a concept!) might
> be a special issue of IEEE TEC on coevolution. Any
> volunteers for guest editors?
I am reluctant to take up another service project, as I already have
my hands full with FOGA ... and that's over a year away. But *AM*
interested in discussing the idea of a special issue, even if I am not
willing to edit it. 8^)
Whenever someone mentions the possibility of a special issue, I
immediately think: Do I have anything to submit to such an issue?
And, of course, I think I do. But my work is (typically) theoretical
in nature and I have a relatively oversimplified hash in my mind
placing theory in the TCS/ECJ bucket and more applied, empirical work
in the IEEE bucket.
Moreover, one of my major concerns is not just raising the scope of
CoEC, but in particular raising the scope of the newer perspectives on
CoEC made possible by recent theoretical advances. I'm just not sure
IEEE TEC is the right venue for that.
Am I being unfair?
Would it be unfair of me to use this IEEE Google Group to posit other
possibilities? For example, the new Natural Computating journal, or
even ECJ. Better yet: I wonder if there's a chance we could get a
special issue in TCSC?
Additionally, a general survey, of the four people that regularly read
this: How many would have something they would want to submit to a
special issue in the near future?
Thanks,
Paul.
R. Paul Wiegand
blithely throwing about.
TCSC -- Journal of Theoretical Computer Science, Section C (Natural
Computing)
[Elsevier: http://ees.elsevier.com/tcs/ ]
ECJ -- Evolutionary Computation Journal
[MIT Press: http://www.mitpressjournals.org/loi/evco ]
TEC -- IEEE Transactions of Evolutionary Compuation
[IEEE Press: http://ieeexplore.ieee.org/xpl/RecentIssue.jsp?punumber=4235
]
NatComp -- Journal of Natural Computing
[Springer: http://www.springerlink.com/content/108905/ ]
Also, here's an incomplete list of AI journals:
http://www.lsi.upc.es/~miquel/aijournals.html
Anthony Bucci
Paul, are you involved with organizing the next FOGA?
Anthony
phi
not sufficiently engaged in CoEC research to offer up a quality
contribution, though I'd be interested in working with someone on some
kind of survey paper, perhaps on games, or on multi-objective. I would
hope that others, not necessarily in this group, would have plenty to
say.
Are you being unfair? Yes, I think maybe so! It's true that in
general, TEC seems to lean towards empirical work, but not exclusively
so. And personally, I'd like to see more theory papers published there
and a greater awareness of theory shown in all papers. Such a change
would have to start somewhere. Why not with this special issue?
phi
R. Paul Wiegand
I was only joking that there seems to be a small number of people
contributing to the discussion in the Google group. (You are involved
in the discussions so you would be one of the "four" to which I was
referring 8^) ).
> Are you being unfair? Yes, I think maybe so! It's true that in
> general, TEC seems to lean towards empirical work, but not exclusively
> so. And personally, I'd like to see more theory papers published there
> and a greater awareness of theory shown in all papers. Such a change
> would have to start somewhere. Why not with this special issue?
Fair enough. I'll give it some thought. Of course, it may in large
part depend on who picks up the baton and pushes for a special issue.
Of course, there's an advantage in TCSC, as well ... a wider audience,
a pretty strong level of respectability in the CS community in
general. I'm not saying TEC is not respectable, just that there may
be (perhaps) a broader community in which TCSC is respected. But all
this is speculation.
Anthony asked: "Paul, are you involved with organizing the next FOGA?
"
(A shameless plug:) Yes. The next FOGA will in be January 2009 in
Orlando. Annie Wu, Thomas Jansen, Ivan Garibay, and I are among those
putting it together. Web page and such coming soon ...
Paul.
se...@eecs.harvard.edu
It hadn't occurred to me to characterize ECJ and TEC as theory and
applications journals, respectively. Looking over their contents in a
cursory way, perhaps this trend does exist somewhat. But, I agree with
Phi that there's plenty to theory published in TEC; let's not forget
that the NFL papers are both published there! And, TEC consistently
has very high impact factors.
As for a special issue on coevolution, TEC did have a special issue on
games 9/6 (2005), which did have a few papers about coevolution; yet
other papers used coevolution, but were about other things, like
evolution of coooperation. So perhaps that 2005 special issue doesn't
make a new coevolution special issue inappropriate. I would be able to
contribute something to such a special issue, too. I'm open to co-
editing, also, conditioned on things like when we plan to do this,
etc..
With respect to TCS, if we want "modern" coevolutionary work to gain
wider acceptance in the EC community, it might be that TCS is too
remote in topic to help with that. One can argue that it will help
promote coevolution to a wider audience, however. I guess we need to
decide who we want our target audience to be, and the purpose behind
selecting that target audience. If we decide to focus on the EC
community, to begin with, then perhaps we want an EC journal (i.e.,
ECJ or TEC). I'd make a similar argument about Natural Computing -- we
need to decide who we want our audience to be and why, and then go
after the appropriate journal. Also, consider that if propose a
special issue for TCS, for example, who of their regular readership is
going to submit? To get papers, we're going to have to advertise in
the EC community, probably; the set of contributors may be disjoint
from the set of regular readers.
To summarize:
1) other threads and discussions have expressed a frustration that,
while a) coevolution is viewed as being important, and b) coevolution
events (tracks, tutorials) are consistently well attended, we seem to
have a problem A) bringing new people into the field, B) getting new
ideas adopted by the community of people using coevolution.
2) to help with this problem, ideas that have been proposed include a)
more tutorials, b) competitions, c) special issues, d) special
coevolution-focused workshops outside of CEC and gecco, e) killer apps
for the new ideas, and so on.
3) if we want coevolution (and modern coevolution) ideas to get more
widely adopted, and we want to do a special issue, what community of
researchers do we want to begin with? Do we want to start with A)
people using coevolution, but with older methods? B) EC people
generally, who at least have heard of coevolution, may think it is
important, and likely know that we are active in coevolution? C)
people more outside of EC, who come to coev tabula rasa and don't have
biases towards older methods?
Also, to promote the modern coevolution ideas, it seems like we'll
have to be the ones to do that; so, a special issue without (or with
few) contributions from us will certainly help promote coevolution,
but it might not advance the state of acceptance of the newer ideas.
Yet, a special issue with mostly contributions from us might be a bit
insular.
-Sevan
R. Paul Wiegand
If no one else will agree to help Sevan co-edit a special issue, I
would do so (with similar conditions as Sevan notes); however, it is
spreading myself a bit thin ... so I'll happily yield that honor to
someone else who wants to "step up"?
As to point number three, the ideal would be to attract some
distribution encompassing all three categories Sevan mentions, but I
don't think that will be possible. The crossover appeal of each of
these journals is probably quite limited ... though I speculate Nat.
Comp. might have a fairly interesting demographic in terms of
interest.
What if we ask ourselves a slightly different question: Which of the
three categories offers the most potential in terms of numbers of eyes
looking at the papers and likelihood to draw interest? Without reason
or evidence, I guess (B). The more narrow group of coev people is
probably too small to expect much, and the broader group probably will
not look closely at the papers due to lack of sufficient interest.
That tilts things toward ECJ, TEC, or GPEM (sorry, I forgot to list
GPEM in my previous message).
Maybe this is a silly choice to debate before approaching various
journals. Some journals may not be interested in such a special
issue, and others may have time-lines that don't work for us. Perhaps
we could produce a short-list, iron out some high-level details (time-
frame, specific topic), then approach the journals on our short list
for interest? Constraints may remove choices.
As to the inbreeding problem, Sevan is bang-on. This is a real
problem, I think. I brought up the issue of having something to
submit only to determine whether there would even be sufficient
content to put an issue out in the first place. Obviously, it would
be good to have as many submissions to such an issue as is possible,
from as many different sources as possible. That might inform our
topic ... a more focused topic would be useful from an advocacy point
of view (i.e., modern coevolutionary foundations), but it would
certainly narrow the pool of potential submissions. On the other
hand, a broader topic (e.g., "coevolution") might draw a lot more
people but might run the risk of diluting a good opportunity.
Would it be over the top to mine the last few conference proceedings
for coev researchers, then email them to see if they would be
interested in submitting to such a special issue?
Paul.
abucci
done; the other is to advance the limits of what is known. I tend to
think that workshops, tutorials, discussion groups, courses, and books
are better venues for disseminating ideas than a special issue of a
journal would be. Journals are where you put research results. A
book which found its way into the hands of new researchers or was used
to teach special topics in courses would spread the word more
reliably, wouldn't it?
It's true that a result with a high "wow" factor published in a
conference or journal can bring a lot of attention to the field. But
to the extent that such results are paradigm shifts, we cannot sit
around waiting for one to appear or expect that our own work is going
to result in one any time soon (it's not easy to predict what's going
to capture people's attention, in any case). In the meantime, I think
persistence is the only strategy, so we continue to hold and improve
tutorials and other such events. We consistently have good attendance
at these, and if we're persistent enough the ideas are bound to
spread, aren't they? It may be frustrating, but so is sanding wood.
I've spent a bunch of time thinking and talking about why the average
EC person is stuck in 1998 when it comes to thinking about
coevolution, but I still have no idea why. So I keep thinking that we
should adjust the tutorials each year to try out new ways of
presenting the information until we find the magic formula.
Since special journal issues on coevolution are so few and far
between, I personally would like to send some new (read: unpublished)
results to one. After all, those of us actively researching
coevolution need a venue for publishing new work, don't we? I don't
think there's anything wrong with the dialog among active researchers
outpacing what the community at large can easily follow. I suppose
there's a fine line between insular and communal, but the field will
stop going anywhere if the people who actively advance it spend all
their time trying to get everyone else to understand what was done
five years ago.
I'd happily co-edit, or even guest edit outright, a special issue on
coev.
Sorry about beating a dead horse with so many words. I wonder if
Jordan would be interested in doing that.
Anthony
abucci
Jordan asking if he'd be interested in editing a journal issue on
coev. T Awhile back I'd emailed Marc Schoenauer about having a
special issue of ECJ but he never wrote me back. I assembled a list
of potential contributors and started into writing a CFP, so maybe we
can use some of that.
Sevan G. Ficici
I agree that a book might be a better way to spread the word than a
special issue, but at least the special issue might be a way to
stimulate participation in coevolution research.
I agree that persistence is necessary to get the word out -- it will
take a little time. I think older coevolution heuristics are more
commonly used for a few reasons: 1) they've been around longer and so
have had more time to be picked up by researchers, 2) they are
typically quite simple to implement, 3) several (e.g., Rosin's)
heuristics have been demonstrated to be helpful in many circumstances.
On the other side, more recent approaches are beginning to get picked
up, as well. Heuristics like LAPCA and the Nash memory have been used
by other researchers, for example.
-Sevan
Anthony Bucci
He also pointed out that it doesn't look good if most of the papers in the issue are written by the guest editor(s).
I'm wondering about turning the question around and seeing what other people are interested in writing about coevolution. We all have our preferences about what we'd like to write or read, but what about other people, the ones we're hoping to reach? What would they want to submit?
Anthony
Hi All,
I agree that a book might be a better way to spread the word than a
special issue, but at least the special issue might be a way to
stimulate participation in coevolution research.
I agree that persistence is necessary to get the word out -- it will
take a little time. I think older coevolution heuristics are more
commonly used for a few reasons: 1) they've been around longer and so
have had more time to be picked up by researchers, 2) they are
typically quite simple to implement, 3) several ( e.g., Rosin's)
R. Paul Wiegand
group. What's the best way to go about that? Have you a list of
people in mind? We could scan over the last few conference
proceedings and email authors of coevolution papers. Or are you
wondering about the wider EC audience?
As to many guest editors problem: I actually don't <i>want</i> to
guest edit, I'm just willing to help if there's no one else. It
sounds like that's not a problem ... so that's one less to worry about
on that front.
Sevan G. Ficici
to contribute to a special issue -- that might be helpful in selling
the idea to the journal. But, I'm a bit unclear about the purpose of
seeing what topics in coevolution others want to publish about.
Wouldn't we want to open the special issue to a broad range of
possible topics? That is, theory, applications, new algorithms, and
so on. It's great to learn about what people are interested in, but
I'm unclear how this information affects our decision making process
(because I assume we'll have a broad CFP -- maybe that's not what you
all have in mind?).
-Sevan
Anthony Bucci
That's why I suggested poking around to get a sense for what other people would like to see in a special issue. There have been at least two applications of LAPCA to "real" problems in the last couple years, for instance. That's the sort of middle ground I had in mind.
Personally, I'd like to see stuff like:
-- replication of older results using newer algorithms
-- comparisons of several methods on the same (set of) problem(s)
-- analysis of cost/performance metrics for algorithms
-- anything about representation in coevolutionary algorithms
-- how coev scales, particularly questions about building blocks/modularization
-- coevolution and robustness
I don't think we'll get submissions focused around topics like this if the CFP is too broad. And I also don't know if that list only reflects my own set of preferences or if it reflects what a lot of people would be interested in seeing in a special issue.
Anthony