Call for papers: Genetic Programming Theory and Practice 2007

The Center for the Study of Complex Systems (CSCS) at the University of Michigan is pleased to be hosting:

GPTP-2007 -- Fifth Annual Genetic Programming Theory and Practice Workshop
May 17-19 (Thur-Sat), 2007
Ann Arbor Michigan USA

GPTP is a small, one-track, invitation-only workshop devoted to the integration of GP theory and practice. In particular, it focuses on how theory can inform practice and what practice reveals about theory. Past workshops have invited speakers to discuss theoretical work and its value to practitioners of the art, and to review problems and observations from practice that challenge existing theory.

This year we are asking researchers who are interested in being invited to present a paper at this Workshop to submit an abstract representing the work they would like to present. We will extend a limited number of invitations to participate based on these abstracts.

We especially encourage papers which combine theory with results from real-world applications, e.g., papers in which theory is used to guide application to real-world problems, and/or results from real-world applications that confirm (or contradict) theory. In this vein, collaborations between "theorists" and "practitioners" are most welcome, since such collaborations are quite rare but they are likely to lead to new insights and novel results.

Some of the questions this workshop is expected to address include:

Does the schema (or other analogous) theorem hold for GP, and what does this mean for the design and use of GP applications?
What are good ways to approach the application of GP to a new problem? What are good ways to adjust control parameters within GP? Are there heuristics to guide the choice of representation, search operators, and so on?
What are the "symptoms" of GP pathology (i.e., premature convergence, overfitting, etc.)?
How universal is the behavior of GP over different problems and data types?
What are the good and bad features of GP for problem application?
Are there ways to predict whether a problem is "GP tractable?"
Application of GP to "real" problems, e.g., applications with large amounts of noisty data, or applications in which GP must compete with other more accepted approaches.

Note that while debate between theorists is encouraged, the focus should remain on the implications for GP applications. Similarly, while "war stories" from application engineers are useful, the choice of war stories should be made on the basis of understanding the GP mechanisms, either by addressing why a particular approach worked or proposing a reason for an unusual behavior.

To see a list of participants and papers for previous GPTP workshops, visit

http://cscs.umich.edu/events/gptp-workshops

Papers have been published in a series of Genetic Programming Theory and Practice books, one for each year (by Kluwer/Springer).

The format of this workshop will be similar to that of the previous workshops (GPTP 2003-2006): a relatively small, invitation-only workshop on the campus of the University of Michigan in Ann Arbor, with plenty of time for discussion of a roughly equal mix of a total of 18 papers by theorists and practitioners. In order to facilitate a substantial exchange of ideas, workshop talks will be approximately 30 minutes long, with considerable time allocated for discussion. The papers will be reviewed by co-participants prior to the workshop, and collected for publication in a book to be published as soon as possible after the workshop. As we would like to have the papers available for distribution before the workshop, the likely deadline for the submission of papers will be mid/late March 2007. Papers will be about 16 pages long.

If you are interested in having a paper considered for presentation at the workshop and included in the book, please send a one-page abstract as well as a short CV of the authors to:

gptp-2007@umich.edu

by: 7 January 2007 *** Earlier responses are encouraged ***

The abstracts will be reviewed and decisions made by 14 January 2007. We will select 6-8 proposed papers based on relevance to the GPTP workshop goals, the expected quality of the contribution, and how the paper topic will fit with the "mix" of other invited and selected papers.

Note that we will be selecting only high-quality, high-content papers, i.e., we are looking for papers that have more depth and breadth than typical conference papers.

Also note there is no "workshop fee" for participation -- the workshop is funded by donations from groups and companies interested in advancing the art and science of genetic programming.

If you do anticipate sending an abstract, please let us know as soon as possible, just so we have a rough idea of how many abstract to expect.

If you have questions, please email them to gptp-2007@umich.edu .

GPTP-2007 Workshop Organization Committee

Rick Riolo
Terry Soule
Bill Worzel