tisk | prohledávání 
 
 
zpravodaj ČSKI
aktuální číslo
přednáška
archiv
  - 2013
  - 2012
  - 2011
  - 2010
  - 2009
  - 2008
  - 2007
  - 2006
  - 2005
  - 2004
  - 2003
  - 2002
  - 2001
  - 2000
  - 1999
  - 1998
  - 1997
  - 1996
  - 1995
  - 1994
 
 
zpravodaj ČSKI - duben 2000 [ pdf ]
[ zaslat oznámení o přednášce ]
leden|únor|březen|duben|květen|červen|říjen|listopad|prosinec


datum: 26.4.2000 v 15:00
název: Imprecise probabilities: discussion of the basic ideas
přednášející: Gert de Cooman (Universiteit Gent, Belgie)
místo konání: VSE, klub knihovny

souhrn: Imprecise probability theory is a generalisation of both probability theory and classical logic, which allows for modelling ignorance as well as uncertainty in knowledge representation and reasoning. The talk will explore and explain the basic ideas behind the theory, and will then highlight its advantages and achievements.
The theory is built upon three pillars: an operational definition of the basic tools for knowledge and belief representation (lower and upper prevision/probability) in terms of behaviour, the basic rationality criteria of avoiding sure loss and coherence these representations should satisfy, and a very powerful reasoning mechanism called natural extension. In the first part of the talk, all these notions will be explained and motivated using simple examples.
In the second part of the talk, it will be argued why anyone should want to use this model: three reasons that will be explored are its generality, its practicality, and its unifying power. It will be shown that (i) most uncertainty models extant in the literature fit nicely into the general framework of imprecise probabilities; (ii) that the knowledge representation and reasoning methods which the theory provides, and which are justified by and based on the three above-mentioned pillars, are particularly easy to implement using straightforward linear programming algorithms; and (iii) that the theory provides criteria for tidying up certain areas of uncertainty modelling, and at the same time gives methods for making them more powerful from a practical point of view. A gain, these points will be explained using simple examples.