************************************************************************
THE SECOND BAYESIAN MODELING APPLICATIONS WORKSHOP
DURING UAI-04
CALL FOR PARTICIPATION
July 7th, 2004
Banff, Canada
Submission of statements due: April 20th, 2004
************************************************************************
GOAL
----
As in the previous year, the goal of the workshop is to encourage
practitioners and tool developers to exchange expertise, techniques,
and methodology in the application of Bayesian methods to a wide range
of application areas and to share domain specific insights in a
focused, informal forum. As an additional avenue of exchange,
practitioners will also have the opportunity to submit statements,
meeting notes, code and examples to be compiled into proceedings as
part of the overall conference publication.
MOTIVATION
----------
Since the first UAI conference, Bayesian modeling methods have matured
and developed, and have seen widespread application. This has led to
local concentrations of experience and know-how. Due to the unique
character of applied work, exchange of information among practitioners
has been difficult. The reasons include:
1. Applications' solutions are often blends of related
techniques. The challenge is often to assemble a coherent
approach that is responsive to the problem.
2. Applications work requires a wide span of knowledge, and this
breadth of knowledge can come at the expense of the depth
required for basic research. As a result, practitioners often
do not participate in highly technical meetings.
3. Insights from applied work often do not fit well into conventional
peer-reviewed paper submissions. They may appear better in
verbal exchange, demonstration, or data presentation formats.
4. Practitioners are often constrained by review and confidentiality
concerns that apply to published material, but not to more
informal discussion. Typically, successful models are more the
subjects of intellectual property protection, whereas there is
little motivation to report unsuccessful results.
As the popularity of Bayesian methods increases and more users with
limited theoretical experience venture to apply existing tools to
diverse and complex domains, the demand grows for sound methodology in
model building. Model builders face questions of model building costs,
model performance and scalability, and validation of the theoretical
guaranties. Moreover, as industry trends continue towards increased
automation, efficiency and productivity, more attention will be paid
to the ability of technology to deliver on its claimed guaranties.
We are looking for participants who are versed in current Bayesian
methods and also have direct experience in problem solving for
specific domains. We expect that the interplay of theory and
application will lead to insights into what works and
why. Participants must be able to describe their experience applying
their tools or methods and offer insights and lessons learned for
their particular application. Special attention will be given to
customer and user issues, such as how to balance the "pull" from
customer needs versus the "push" from technology.
Developers of tools -- for both modeling and elicitation -- are also
welcome to demonstrate and discuss their applications.
SUGGESTED TOPICS AREAS
------
This year, instead of defining subject areas by domain, we emphasize
various aspects of modeling we want participants to address:
- How successful projects are initiated; how to identify a fruitful
opportunity and sophisticated client.
- How judgment elicitation techniques can be assisted by data-driven
methods and vice versa.
- How modeling techniques may vary depending on the domain, maturity
of the application, or the organization.
- What other applied fields should be drawn on to complement modeling?
- How applications are justified, formulated, validated, fielded, and
maintained.
- Interactions between academic research and problem solving: What are
effective models of collaboration.
While the emphasis is in methodology we are encouraging submissions
that cover a broad range of areas such as diagnosis, optimization,
learning, temporal reasoning, speech and visual feature recognition,
robotic navigation, language and document analysis. Novel applications
in fields such as biology and economics are also encouraged.
FORMAT
------
The workshop will take place on Wednesday, July 7th, the day before
the UAI tutorials. It will consist of 5 interactive sessions of groups
of participants. Composition of the groups will be selected by the
organizing committee to maximize the possibility of interesting and
fruitful discussion within the group. Each workshop session will be
devoted to one group. The workshop will begin with highlights
presentations where each participant will give a short introduction to
themselves and their work. These brief introductions will be followed
by the sessions, consisting of open discussion among group members and
the audience on issues related to the areas of interest of the group.
One of workshop organizer will moderate each session.
Of special interest this year will be a planned final session
involving invited clients who can speak to their experience
introducing applications into their organizations. Participants are
encouraged to contact organizers with individuals who can attend in
this role.
ARRANGEMENTS
------------
The workshop will run on the premises of the general UAI conference.
The registration fee of $40 for the workshop will be collected
separately from the main conference. Arrangements will be made to
collect fees before the day of the workshop.
TO PARTICIPATE
--------------
Persons desiring to participate should prepare a short (three pages)
statement of interest, explaining their background, the applications
in which they have contributed and the characteristics of the domain
in which they work. The statement should raise questions and offer
results, which the participant can speak about. All participants will
be expected to have material to present, either formal or informal,
for the workshop. Participants are encouraged to apply jointly with
members of other disciplines with whom they have collaborated. Make
all submissions and enquires to [EMAIL PROTECTED]
Participation will be limited to 30 individuals. Acceptance will be
biased in favor of statements that touch upon modeling methodology and
customer issues. We strongly encourage submissions to identify the
customers and to discuss how their needs influenced the solution. All
members of the workshop committee who would like to attend are
expected to qualify as participants. Selection will be done by the
committee's decision of the mix of individuals that are likely to
generate the most constructive presentations and discussions.
ORGANIZING COMMITTEE
--------------------
Oscar Kipersztok, The Boeing Company, Chair
John Mark Agosta, Intel Corporation, Co-chair
Russell Almond, Educational Testing Service
Bert Kappen, University of Nijmegen
Kathryn Blackmond Laskey, George Mason University
K. Wojtek Przytula, HRL Laboratories
Irina Rish, IBM T.J. Watson Research Center
IMPORTANT DATES
---------------
Submission of Statements of Interest: April 20th.
Notification of selections by the organizing committee: June 1st.
Deadline for optional contributions to workshop proceedings: June 18th
