Skip to main content

Defeasible Reasoning and Uncertainty Management Systems

Objective

DRUMS aimed to study the integration of several forms of non-standard logics and models that could be applied to problems of defeasible reasoning and uncertainty management. The models studied are necessary for automated reasoning at the level of:
-knowledge representation where knowledge is pervaded with some form of ignorance, be it incompleteness, imprecision or uncertainty
-inference engines adapted to each form of ignorance.
The final aim of DRUMS was to attempt the integration of various logics and models into a general model able to cope with the various forms of ignorance, and its implementation in inference engines.

APPROACH AND METHODS
DRUMS was realised through an interdisciplinary and collaborative approach covering the following four research domains:
-Logic models for defeasible reasoning, covering default logics, the integration of concurrent representations, and proof procedures and algorithms.
-Probabilistic models of reasoning under uncertainty and vagueness, addressing the development of possibilistic and linguistic quantifier logics and knowledge elicitation.
-Models of belief for approximate reasoning, including the transferable belief model, approximations and algorithms, and propagating uncertainty in dependence graphs.
-Combined models of reasoning with uncertainty, covering an extended logic language and advanced prototyping.
PROGRESS AND RESULTS
Scientific papers have been published and presented at several congresses. They cover each of the aforementioned domains of research and several integration issues. Six workshops and an International Conference have been organised. Several pieces of software dealing with the management of uncertainty have been successfully developed.
POTENTIAL
The improvement of expert system shells on the aspect of uncertainty management is very important for the knowledge-engineering industry. So far, only ad hoc and poorly justified approaches have been available to users, leading to poor results and dissatisfaction. Indeed, for most expert systems developed, the need for uncertainty and incompleteness has not been paramount because of their size and simplicity. As systems become more complex, however, uncertainty and incompleteness become crucial in determining the choices to be made. Furthermore, a proper treatment of uncertainty would improve the behaviour of expert systems.

Coordinator

VRIJE UNIVERSITEIT BRUSSEL
Address
Avenue F.d. Roosevelt, 50, 122
1050 Bruxelles
Belgium

Participants (8)

CENTRE D'ESTUDIS AVANCAIS DE BLANES
Spain
Address
Cami De Santa Barbara
17300 Blanes
Centre National de la Recherche Scientifique (CNRS)
France
Address
Campus Du Beaulieu
35042 Rennes
Centre National de la Recherche Scientifique (CNRS)
France
Address
118 Route De Narbonne
31062 Toulouse
Imperial Cancer Research Fund (ICRF)
United Kingdom
Address
44 Lincoln's Inn Fields
WC2A 3PX London
Queen Mary and Westfield College
United Kingdom
Address
Mile End Road
E1 4NS London
UNIVERSITAT DE GRANADA
Spain
Address
Fuentenueva
18071 Granada
Université d'Aix-Marseille III (Université de Droit d'Économie et des Sciences)
France
Address
70 Route Léon Lachamp
13288 Marseille
Université de Paris XI (Université Paris-Sud)
France
Address
Avenue Georges Clémenceau
91405 Orsay