Oesterreichisches Forschungsinstitut fuer Artificial Intelligence (OeFAI)
Schottengasse 3, A-1010 Wien
Tel.+43-1-53532810,5336112
-------------------------------------------------------------------------
VORTRAG
*******
Prof.Dr.Yves KODRATOFF
Directeur de recherche au CNRS
Orsay
France
The Comprehensibility Manifesto
or
"Comprehensibility in View of Industrial Applications"
------------------------------------------------------
When I had to prepare a talk on industrial applications of Machine
Learning, I wanted to underline the main result that showed the deep
commonality among all these real world applications, how industrial
constraints would force the academics to think differently. To my
surprise, some applications request very simple algorithms, others
use the most complex multistrategy approaches, some have plenty of
noise, for some others the industrial constraint is to forbid any
kind of noise, etc. In other words, industrial constraints are varied
to an unbelievable point. Only one thing is in common to all industrial
applications of Machine Learning: the users demanded comprehensibility
of the results of the Machine Learning algorithms, and the greater the
comprehensibility, all the more enthusiastic they have been. This is
why I suggest to stop giving absolute supremacy to measuring the
complexity of the algorithms and the accuracy of their results. Let us
be concerned also by the comprehensibility of the software and of the
results. The problem is that we do not understand comprehensibility!
This is why I propose to stop fleeing from the problem and define
comprehensibility as an acknowledged research topic.
I will discuss some of the interdisciplinary issues and also some of the
pure computer science problems linked to comprehensibility.
Zeit: Dienstag, 26.Juli 1994, 18:30 Uhr pktl.
Ort: OeFAI, Schottengasse 3, 1010 Wien 1.