Wednesday, June 13, 2007

Broad coverage of topics

This is perhaps one of the most important issues that our community faces. A broad coverage should facilitate freshness and innovation along with continuity. It should allow for novel problems and applications, and recognize new ideas or techniques originating in other parts of mathematics and computer science that address old open questions.

The community must be open and recognize both the results emerging from larger-scale, well-funded research projects, as well as those coming from individual, perhaps more isolated researchers. A broad coverage may be stiffled by the effect of establishment, connections, name recognition or coasting on reputation. A broad coverage should allow co-existence of fundamental results with hot, application-oriented ones. A broad coverage should invite participation in our conferences of
members of other communities whose problems and techniques are geometrical.

What should we do to facilitate/ensure a broad coverage of topics?


Frederic said...

I believe Computational Geometry should be taken in a rather eclectic
sense: anything involving algorithms and calculations on geometric
objects, with a mathematical inclination, should fall in its realm. a
recent initiative in this spirit is certainly the Symposium on
Geometry Processing, as one finds there attendees from classical CG,
CAGD and Computer Graphics/Geometry Processing. to broaden the scope
of SoCG, I believe two things would help a lot:

1. a better and more open review process --see the thread on Best
Practices, meant to serve an expanded sense for 'Computational

2. an acknowledgment of the fact that privileged ways to disseminate
geometric ideas and concepts in other communities are quality software
and applied papers. unfortunately, papers dealing with complete and
robust implementations often look tedious to people essentially
interested in generic situations and complexity issues, so that
getting such papers accepted gets more and more difficult. cf also the
CG Task force.

Herman said...

Accept more papers (facilitated by making the talks shorter and having more parallel sessions), and thus, more people.

Anonymous said...

I think "broad coverage of topics" is definitely a goal that we should strive for.
The May 2007 issue of Communications of the ACM contains an article on the history of SIGGRAPH:

"[The SIGGRAPH conference] was also the place to see everyone with whom you wanted to talk about computer graphics." (p.57)

SoCG should ideally be the conference where everybody who is attached to computational geometry and has no conflicting obligations would regularly want to go, to hear the latest results and keep up-to-date, to present their results, and to meet and talk to people.

This time, not even all authors of accepted papers were present (which might partly due to the remote location for many people and to the competition of the adjacent Japanese conference).

I think we could accept more papers without compromising the quality of the conference, as viewed by how interesting it is for the participants.
My experience on program committees was that there was in the end a tough decision about some remaining papers that could have gone either way, and we could have accepted all of these had it not been for limited space in the proceedings and limited time at the conference. In this final stage, it is more likely for papers that are not "core" computational geometry, that are at the fringe, to get rejected.

This is, sadly, in conflict with the explicit goal of encouraging "applications", "software", "experimental" papers, and papers from related fields.

We should try to explore different formats which allow more people to participate, and at the same time keep the conference competitive.
I would personally be in favor of reviving the poster session, for example.

Releasing pressure in this direction, in one way or another, might make the discussion about "consistency of standards" or "accountability for one's judgements" obsolete.
(If a paper is rejected for no better reason than that there were other papers that were considered better, what would be "proper feedback" in this case?)
Günter Rote

Joseph Malkevitch said...

Now that computational geometry is maturing as a field it would be
nice if research practitioners paid some attention to educational and
public awareness issues. Many topics and accomplishments in
computational geometry should be things that high school and
undergraduate mathematics teachers and undergraduate CS teachers
should know about. (The ultimate goal being public awareness in
general.) Especially when there is work that results in applications
that affect people in everyday life this information should be made
available to teachers. Thus, computational geometry that helps
improve cell phone service, GPS, etc. should be promoted as a
consequence of supporting research in Computational Geometry.

Frederic said...

a few additional comments:

- at Siggraph, there is no predefined number of papers to be accepted:
the quality of submissions sets the pace, and the organizers manage.

- I do not think posters are the solution : a poster is not a full
publication, and people are unlikely to travel / spend time+money for
a jump seat.

- an important reason to attend a conference is to get exposed to
stimulating and new topics. I do not think novelty will be boosted as
long as the distinction between 'core CG' and 'non core CG' will
remain. cf also SGP, where a number of 'core CGers' go, because it's a
nice forum to get exposed to computational geometry in a more
eclectic sense.

Anonymous said...

Frederic says
" I do not think novelty will be boosted as long as the distinction between 'core CG' and 'non core CG' will remain. "

I tend to agree on this.
In fact I don't really know what 'core CG' is.
I see several different cores in CG... And there are surely many potential interesting geometric topics outside these cores.

Another remark: like Siggraph, SoCG has also no predefined number of accepted papers. Even though the final number is rather stable through years.

I already see SoCG as the conference (maybe not the only one) I would like to attend regularly. Still, we don't have a budget that is large enough to send all members of the group every year... SoCG cannot find a solution to this practical question.

Monique Teillaud

Anonymous said...

Another remark: like Siggraph, SoCG has also no predefined number of accepted papers.

Again this is not so. In certain cases the venue and dates selected imposes a hard limit on the number of papers. It is my understanding that this has been an issue at least twice in the last ten years or so.

Anonymous said...

About so-called "core" and "non-core" Computational Geometry.

Of course, there is no clear-cut separation here, but one might define the "core" of Computational Geometry, as the subjects that form traditionally the main part of SoCG, or subjects that many members of the SoCG program committe are familiar with, and would feel comfortable to review.

For example, a paper about Voronoi diagrams or Delaunay triangulations it most probably core CG; a paper about planar graphs with graph-theoretic techniques, with has some connection to geometry, is "non-core" CG. Anything radically new is by definition "non-core" CG.

This is not meant to exclude "non-core" CG from SoCG. But even if one does not like it, this distinction exists, and it will naturally tend to reflect itself in the decisions of the committee, and unless one makes a conscious effort to fight it, it will tend to perpetuate itself, and make it hard for new trends or new ideas to become established.
Günter Rote

Anonymous said...

Guenter wrote: [...]
one might define the "core" of Computational Geometry, as the subjects that form traditionally the main part of SoCG, or subjects that many members of the SoCG program committe are familiar with, and would feel comfortable to review.

If it were `the core of SoCG topics'
instead of `the core of computational geometry', this would be a tautology.
But as stated, it is not true, and I think, the entire idea of `Broad Coverage' has a problem there. Because there are many `SoCG topics' which are not computational geometry, but got in because of an overlap of interests in the persons of some traditional members of the SoCG community. Some combinatorial geometry got in, and a subset of discrete geometry essentially defined by being a side interest of a
core SoCG member, e.g. recently transversal questions. This leads to the illusion that SoCG is really the conference of Discrete and Computational Geometry, but it is not, there is a big Discrete Geometry community out there which is not related to the SoCG community. E.g. the community of the hungarian `Intuitive Geometry' conferences, as long as they existed, is, with the exception of the combinatorial geometry part, disjoint to the SoCG community. The polyhedra people are disjoint. And there are a number of other groups and topics, whose inclusion or exclusion in SoCG depends on whether there is a core SoCG member that also works on that topic.

Thus the SoCG is automatically a meeting defined not only by topic, but to a certain degree by group membership. If the conference would restrict itself to really core computational geometry, this problem would not exist, but exactly the attempt to `broaden the coverage' leads to a broadening defined by group membership. And that leads to disappointed expectation in those who are outside the core, but think that they are not more distant to the core than others who did get accepted.

So I think that a more restrictive view of the topic, aiming to accept all competent papers on computational geometry, but rejecting the papers outside the topic, would make the decision process less controversial.

Peter Brass