The Department of Computer Science
			and the research group
	  Declarative Languages and Artificial Intelligence
	    of the Katholieke Universiteit Leuven, Belgium

			      announces


	   Symposium "From Machine Learning to Data Mining"


	  http://www.cs.kuleuven.ac.be/~wimv/PhD/Symposium/
 
		Monday, June 3, 2002, Leuven, Belgium


With support from
* FWO scientific network 
   Machine Learning for Data Mining and its Applications
* FWO scientific network 
   Declarative Methods in Computer Science

Attendance is free, but registration is required (May 27 at the latest).

After the Symposium, at 16h30, there is a PhD defence (in Dutch)
by Wim Van Laer:
"From Propositional to First Order Logic in Machine Learning 
 and Data Mining -- Induction of first order rules with ICL"


*************
* PROGRAMME *
*************

Symposium from Machine Learning to Data Mining

9.30  am  Welcome and coffee

9.45  am  Opening of the seminar

10.00 am  Inductive databases : a declarative data mining approach
          Luc De Raedt, Albert-Ludwigs-Universität Freiburg, Germany 

11.00 am  Coffee break

11.20 am  Building and mining the multidimensional HIV data cube
          Elke Van Craenenbroeck and Luc Dehaspe, PharmaDM, Leuven 

12.20 pm  Sandwich lunch

13.40 pm  Descriptive data mining: current issues
          Peter Flach, University of Bristol, UK 

14.40 pm  Coffee break

15.00 pm  Is Combining Classifiers Better than Selecting the Best One?
          Saso Dzeroski, Jozef Stefan Institute, Slovenia

16.00 pm  Coffee break

**

PhD defense Wim Van Laer

16.30 pm  From Propositional to First Order Logic in Machine Learning 
          and Data Mining -- Induction of first order rules with ICL
          Presentation will be in Dutch

The defense is followed by a reception. 
          

****************
* REGISTRATION *
****************

Registration can be done online 
at http://www.cs.kuleuven.ac.be/~wimv/PhD/Symposium/

It is possible to order lunch (sandwiches) at EUR 10 per person.


***********
* ROADMAP *
***********

All activities will take place in auditorium De Molen,
Kasteelpark Arenberg 50 in Heverlee, Leuven.

A detailed description and a map can be found at the website.


*************
* ABSTRACTS *
*************

Available at the website.



* Inductive databases : a declarative data mining approach

  Luc De Raedt, Albert-Ludwigs-Universität Freiburg, Germany


* Building and mining the multidimensional HIV data cube

  Elke Van Craenenbroeck and Luc Dehaspe, PharmaDM, Leuven 


* Descriptive data mining: current issues

  Peter Flach, University of Bristol, UK


* Is Combining Classifiers Better than Selecting the Best One?

  Saso Dzeroski, Jozef Stefan Institute, Slovenia

We empirically evaluate several state-of-the-art methods for constructing
ensembles of classifiers with stacking and show that they perform
(at best) comparably to selecting the best classifier from the ensemble 
by cross validation. We then propose a new method for stacking, 
that uses multi-response model trees at the meta-level, and show that 
it clearly outperforms existing stacking approaches, 
as well as selecting the best classifier from the ensemble 
by cross validation.