May 30TH	Barcelona

(At the European Conference in Machine Learning.)
******************************************************************************
******
                                             ******
******                       ECML'2000 Workshop
                   ******
******        Meta-Learning: Building Automatic Advice Strategies       ******
******            for Model Selection and Method Combination
******
******
                                             ******

******************************************************************************


Motivation and Technical Description
--------------------------------------------------------------------------------
-----------------------------

The application of Machine Learning (ML) and Data Mining (DM) tools to
classification and regression tasks has become a standard, not only in
research but also in administrative agencies, commerce and industry (e.g.,
finance, medicine, engineering). Two important aspects of such application is
the selection of a suitable model and the combination of methods. Since the
expertise to address these issues is seldom available in-house, users of
commercial ML and DM tools must either resort to trial-and-error or
consultation of experts. Clearly, neither solution is completely satisfactory
for the non-expert end-users who wish to access the technology more directly
and cost-effectively. Automatic and systematic guidance is required.

Automatic guidance in model selection and data transformation requires
meta-knowledge. Current ML and DM tools are only as powerful/useful as their
users. One main aim of current research in the community is to develop
meta-learning assistants to support users. Such systems should be able to
deal with the increasing number of models and techniques, and give advice
dynamically on model selection and method combination. Furthermore, inductive
meta-learning capabilities should be included, which use cumulative expertise
gained from prior research and the conclusions of past comparative studies,
all of which are useful forms of meta-knowledge. Meta-learning assistants
could be integrated naturally in future versions of commercial ML and DM tools.


Objective and Scope
--------------------------------------------------------------------------------
----------------------------------------

In the past decade, a number of research projects have examined meta-learning.
In Europe, the most prominent ones include the ESPRIT Statlog (1991-1994) and
METAL (1998-2001) projects. The aim of this workshop is to offer the
international community a forum to exchange experience, knowledge and
perspectives in meta-learning. In particular, it is hoped that academics and
those working in research institutes will present the current status of the
research; ML/DM practitioners will describe applications of meta-learning in
real-world problems; and ML/DM software developers will discuss tools and
potential integration of meta-learning assistants in their systems.

This workshop continues in the tradition of previous related workshops, such
as the ECML95 Workshop on Learning at the Knowledge Level and the ICML97
Workshop on Machine Learning Applications in the Real World. It also
complements the results of the ECML98 Workshop on Upgrading Learning to the
Meta-Level, the AAAI98/ICML98 Workshop on the Methodology of Applying Machine
Learning and the ICML99 Workshop on Advances in Meta-learning and Future
Work.

Contributions (from all main sub-fields of ML) describing work in progress as
well as position papers are invited. All contributions must focus on the
automation of machine learning and meta-learning. Of particular interest are
methods and proposals that address the following issues:

*- What criteria and metrics can be used for evaluating and autmating
          model selection in classification and regression? What is the cost
          of these metrics?
        - How can expert knowledge be integrated with meta-learning?
        - What are the requirements for a dynamic, incremental meta-learning
          system? What multi-criteria advice strategy should be used?
*- What different approaches to meta-learning have been/can be proposed
*  and/or implemented?

Presentations of beta versions of meta-learning tools for automated or guided
use of methods or algorithms with respect to performance or run stability are
also welcome.


Submissions
--------------------------------------------------------------------------------
--------------------------------------

Papers must be submitted electronically, preferably in postscript, to one of
the organisers. Submitted papers will be reviewed by at least two independent
referees from the Program Committee. Accepted papers will be published in the
workshop proceedings and contributors will be allocated 30 minutes for an oral
presentation during the workshop.


Organisation
--------------------------------------------------------------------------------
-----------------------------------------

Organisers:

J. Keller, DaimlerChrysler AG, Germany
[log in to unmask]
C. Giraud-Carrier, University of Bristol, UK             [log in to unmask]

http://www.iiia.csic.es/ecml2000/

Program Committee:

Pavel Brazdil, University of Porto, Portugal
Philip Chan, Florida Institute of Technology, USA
Robert Engels, University of Karlsruhe, Germany
Dieter Fensel, University of Karlsruhe, Germany
Jean-Gabriel Ganascia, Univeriste Paris VI, France
Christophe Giraud-Carrier, University of Bristol, England
Ashok Goel, Georgia Institute of Technology, USA
Melanie Hilario, Univeristy of Geneva, Switzerland
Joerg Keller, DaimlerChrysler AG, Germany
Stan Matwin, University of Ottawa, Canada
Dunja Mladenic, Jozef Stefan Institute, Slovenia
Gholamreza Nakhaeizadeh, Daimler Benz AG, Germany
Bernhard Pfahringer, Austrian Research Institute for AI, Austria
Andreas Prodromidis, Columbia University, USA
Maarten van Someren, University of Amsterdam, The Netherlands
Gerhard Widmer, Austrian Research Institute for AI, Austria
Takahira Yamaguchi, Shizuoka Univeristy, Japan
--------------------------------------------------------------------------------
--------------------------------------------------------------------



--------------------------------------------------------------------------------
--------------------------------------------------------------------------------
---------------------------------------


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%