The report below originally appeared in the IMS Bulletin.
See here or
here
(page 10).
David Banks, Duke University, reports: On May 10-13, the
Classification Society of North America met at DIMACS, the
Center for Discrete Mathematics and Theoretical Computer
Science at Rutgers University. The meeting, which was co-sponsored
by IMS, had a shared day with the DIMACS workshop on
Clustering Problems in Biological Networks.
As is customary at the annual CSNA meetings, the topics
and applications were broad, but the common thread was cluster
analysis and classification. This theme appeared in invited sessions
on social networks, authorship identification, computational biology,
data mining, protein interactions, and weighted clustering.
It also appeared in the keynote talks; Panos Pardalos (University
of Florida) spoke on mining massive datasets, Christina Leslie
(Columbia University) spoke on protein classification, Sal Stolfo
(Columbia University) spoke on anomaly detection in computer
networks, David Madigan (Rutgers University) spoke on use of
domain knowledge for prediction, Sanjoy Dasgupta (University of
California-San Diego) spoke on active learning of linear separators,
Pierre Hansen (GERAD, Montreal) spoke on automated theorem
proving in the context of graph invariants, Casimir Kulikowski
(and collaborators) spoke on cluster analysis in graph-structured
genomic databases, and Fionn Murtagh (University of London)
spoke on ultrametricity.
In terms of take-home messages, the cross-cutting conclusion
is that dynamic network models will play a central role in many
kinds of applications. One key issue is that the appropriate kind of
dynamics depends sensitively upon the scientific context; another is
that assessing the fit and predictive accuracy of such models will be
difficult. A second main message is that there are two broad strategies
for modern classification: tree-based methods and support vector
machines. In applications and test-beds, neither dominates the
other, and it is an open question to determine the circumstances
under which method is likely to be superior. A third main message
is that biological applications are growing fast, but face challenges
both in terms of data quantity and data quality. And a fourth main
message is that the classical tools of statistics are both the foundation
for a new generation of inferential tools and the benchmark
against which success is measured.
CSNA meetings pull together biologists, computer scientists,
psychologists, library scientists, management scientists, mathematicians,
and statisticians. This range of representation ensures that
nearly everyone has some tool or insight to offer to a colleague,
and that nearly everyone can learn something fresh. The CSNA
publishes the Journal of Classification, edited by Willem Heiser
(Leiden University); this journal has a long history of excellence in
publishing leading articles on theory, methods, and applications in
cluster analysis and classification.
This year's meeting was hosted by Mel Janowitz, CSNA president
and associate director of DIMACS. The 2007 meeting will be
at the University of Illinois at Urbana-Champaign.
From the Secretary/Treasurer, Stanley L. Sclove
Department of Information & Decision Sciences,
University of Illinois at Chicago
slsclove at uic.edu
MEETINGS
A very successful CSNA2006 meeting was hosted in May by CSNA President
Mel Janowitz at Rutgers University's Center for Discrete Mathematics
and Computer Science (DIMACS), where Mel is Associate Director.
A report of the meeting is contained in this newsletter.
Dave Dubin is planning to host CSNA2007 at the University of Illinois
at Urbana-Champaign. This will be in early June, in
conjunction with the Digital Humanities group.
Further information is given elsewhere in this newsletter.
Plans for CSNA2008 will begin to be made soon.
Doug Steinley of the University of Missouri has offered to organize
the meeting.
ELECTION AT THE END OF THE YEAR
Each year, the terms of two members of the
Board of Directors expire, and those two positions on the Board are
to be filled. The Nominating Committee
(Herbie Lee, Chair; Mel Janowitz, Fionn Murtagh, Bill Shannon and myself)
has chosen three persons, who have agreed to run. They are
Michael J. Kurtz, Harvard-Smithsonian Center for Astrophysics
Samantha Bates Prins, Department of Statistics, Virginia Tech,
Blacksburg
Padhraic Smyth, Department of Information & Computer Science,
University of California - Irvine
According to the election procedure provided by the by-laws, the
Secretary will issue (by e-mail to the eligible voters)
a call for nominations "from the floor" (by petition of at least five
members). A month is allowed for this. (Often there are no additions
to the slate provided by the Nominating Committee.)
Toward the end of the calendar year, there is a one-month voting period.
The voting will be
on-line by a ballot prepared by CSNA (and IFCS) Webmaster Dave Dubin.
The Secretary will send an e-mail reminder to eligible voters.
MEMBERSHIP
Urge your colleagues to join CSNA !
Benefits of membership include:
A subscription to the Journal of Classification
bibliographic Classification Literature Automated Search Service CD
20% discount on all Springer books
Online access to all back issues of the Journal of Classification
Discount registration fees at the annual CSNA meeting
The 10th Conference, a Jubilee event, of the International
Federation of Classification Societies took place in
Ljubljana, Slovenia, on 25 to 29 July. The venue at the
Faculty of Social Sciences, of the University of Ljubljana,
was most pleasant, and a short distance by bus from the
charming central area of Ljubljana.
CSNA was well represented at the Conference itself, and
also at the IFCS Council meeting which took place on 26
July. Two items of note from the Council meeting:
the Slovene Statistical Society will become a new member
of the IFCS; and the next IFCS Conference will be in
Dresden, Germany, on 17-22 August, 2008.
The opening plenary presentation at IFCS was by Adrian
Raftery, Center for Statistics and the Social Science,
University of Washington, on model-based clustering,
feature selection, and clustering of social networks.
A reception was provided by the Minister of Higher
Education, Science and Technology, in the Government
of Slovenia, Jure Zupan. Dr Zupan is himself author of books
and articles on clustering and data analysis, including
Clustering of Large Data Sets (1982), Algorithms for Chemists
(1989), Neural Networks for Chemists: An Introduction (1993,
Japanese translation, 2000); Neural Networks in Chemistry and
Drug Design (1999); among other works.
At Interface/CSNA 2005 two fantastic presentations were given
that are toappear as Special Invited Papers in this year's Journal of
Classification:
Jon
Kettenring: The Practice of Cluster
Analysis JoC,
Issue 1 of
2006
Jerry
Friedman: Recent Advances in Predictive (Machine)
Learning JoC,
Issue 2 of 2006
I expect these to be well worth having and would
encourage anyone interested to contact Stan Sclove (slsclove at uic
dot edu) about subscriptions.
Interdisciplinarity and Transdisciplinarity in the Organization of
Scientific Knowledge. Organized by the University of Leon.
Leon, Spain, 18 to 20 April 2007. FOR SUBMISSION OF ABSTRACTS AND
GENERAL INFORMATION:
Dra. Blanca Rodríguez Bravo / Dra. María Luisa Alvite Díez.
Biblioteconomía y Documentación Universidad de León.
dphabd@unileon.es,
http://www.ugr.es/~isko/.
ASMDA 2007 International
Conference, May 29 - June 1, 2007, Chania, Crete, Greece
12th International Conference on
Applied Stochastic Models and Data Analysis(ASMDA 2007), which will take
place in Chania, Crete, Greece (May 29 - June 1, 2007). Further
information can be found at the Conference Web site:http://www.asmda.com
NIPS: Call for Workshops and Call for Demonstrations
Neural Information Processing Systems Conference and Workshops
Vancouver and Whistler, BC, Canada December 4-9, 2006.
The Call for Workshops and the Call for Demonstrations are now
available online at: http://www.nips.cc.
The WWW and ascii version of the CSNA Newsletter is made available as
a service of the Classification Society of North America (web site:
http://www.classification-society.org/csna/csna.html ).
Information on becoming a member of
CSNA is available at the CSNA website.
CSNA Webmaster:
Dave Dubin, ddubin at uiuc dot edu
Newsletter editor:
Mike Larsen, larsen at iastate dot edu
