CFP

The study of machine learning and data mining methods might fall prey to the effects of a publication bias that favours sucesses rather than failures. Most conferences and journals papers focus on positive results - either explicitly (because this is a clear acceptance criterion) or implicitly (because researchers have the impression that negative results would not be accepted).

We believe that this policy of favouring success stories does not reflect the practice of a field where failures happen regularly. As a community, we might regard failures as being as informative as successes, for these are our negative examples... It is our hope that this workshop, concentrating on unexpected failures, will provide interesting hints into the current boundaries of our field.

The workshop intends to provide a forum for papers outside the success stream, which is why we called it Salon des refusés. We welcome submission of papers reporting on failures that are significant (as opposed to falling close to the best / average known accuracy) and unexpected with respect to the current state of art and practice.

The discussion will, if at all possible, provide some hints into the failure causes regarding the method and/or the problem characteristics.

Submissions

In order for the workshop to be accessible to a wide audience within the community, we welcome papers that report failures of learning and mining strategies that are already popular and well-known in the community, or of novel ideas that do not require extensive prior knowledge in a micro niche of machine learning.

The mode of failure should be put in understandable terms for the machine learning and data mining practitioner. The causes of failure must be discussed, and if at all possible, explained. As an alternative, complementary (further) experiments might be proposed to get further hints into the failure causes.

The ideal workshop paper should describe an experiment or an idea that is likely to be repeated by other people, and whose expected outcome is clear to everybody. The result should fail this expectation.

The main evaluation criterion will be whether it appears to be worthwhile to record the failure for the community.

Papers should be submitted in electronic form (PDF or PostScript preferred) to all four organizers, using same format as for ECML/PKDD. There is no length restriction at submission time. A maximum length will be set up for the production of the proceedings volume.

All workshops will have the same submission deadlines, which are given below. In order to guarantee a timely production of the proceedings, both the camera-ready copy and a Web-version of tutorial notes and workshop proceedings must be ready by July 12th, 2002.

Important Dates

WS paper submission deadline: 24.05.2002 WS paper acceptance notification: 14.06.2002 WS paper camera-ready deadline: 01.07.2002

Organizers

Hilan Bensusan
http://www.cs.bris.ac.uk/~hilanb/
Department of Philosophy
Universidade de Brasília 04661
Brasilia 70.910-900
Brazil
hilanb@unb.br

Christophe Giraud-Carrier
http://www.cs.bris.ac.uk/~cgc/
ELCA Informatique SA
Av. de la Harpe, 22-24
CH-1000 Lausanne 13
Switzerland
cgc@elca.ch
Phone: + 41 21 613 2111
Fax: +41 21 613 4700

Johannes Fürnkranz
http://www.ai.univie.ac.at/~juffi/
Austrian Research Institute for Artificial Intelligence (ÖFAI)
Schottengasse 3
A-1010 Wien
Austria
juffi@oefai.at
Phone: +43 1 53 36 112 19
Fax: +43 1 53 36 112 77

Michele Sebag
http://www.lri.fr/~sebag/
Laboratoire de Recherche en Informatique
Université Paris-Sud Orsay
F-91405 Orsay
France
sebag@lri.fr
Phone: +33 1 69 15 64 53
Fax: +33 1 69 15 65 86