En continuant à naviguer sur notre site, vous acceptez l'utilisation de cookies pour vous proposer des services et offres adaptés à vos centres d'intérêts.
> En savoir plus

Description Haut de page


 Methods and tools


  Probabilistic networks applied to risk management and dependability


  Philippe WEBER et Christophe SIMON (CRAN)



This WG has been created folowing a working day dedicated to  «  bayesian networks : Methods and applications to risk management and dependability  » organized both by IMdR and RUFEREQ network in september 2012, to share experiences between industials and academics about bayesian networks usage to modelling problems solving  linked to risk management and dependability.

 Bayesian networks are probabilistic networks  considered as a solid mathematical formalism, hold by performing software platforms.  Emergence of ergonomical tools for modelling and probabilistic computation allows a huge industrial and academic community to perform bayesian networks technics, thus disseminating in numerous activity fields as well as application types such as risk analysis, reliability assessment, failure diagnosis, maintenance prognosis, aso...

This mathematic formalism is by now well adopted by the international scientific community. Numerous are researches dealing with computing algorithms development in bigger and bigger probabilistic networks, integrating temporel  variables, continuous variables,...  

Probabilistic networks provide a huge modelling capacity and their computing accuracy is widely demonstrated, namely concerning reliability computation problems. They cannot be ignored to understand complex systems, through scenarii simulation possibility, new knowledge insertion and diagnosis of functionment or dysfunctionment situations causes.

However usage precautions remain, mainly during dynamic processes modelling where temporel processes complexity leads to important size models. Inferential algorithms are continuously enhanced to model more and more complex systems.

In spite of bayesian networks formalism and handling tools maturity which allows their exploitation in industrial and operational problems, approaches, building and structuration model methologies remain to be formalized. : in fact a big effort must be afforded to upstream model  building and knowledge structuration. Model robustness lays on model building methodology relevance and correctness. 

We propose the WG to be focused on pending questions and some perspectives namely about model robustness assessment, dependancies materialization, every kind of uncertainty consideration, opening towards more elaborate and more efficient whose objective is to face huge size modelling problems. Promising formalisms to investigate are Relational Probabilistic Models (RPM), modelling by networks of belief function, aso. These new formalisms are to be studied to master their limits and their potential. We propose to follow the evolution and the new adaptation of modelling tools to answer concrete and more and more complex problems 

This WG gives an opportunity for actors specialists in probabilistic networks model to meet and share  knowledge and experience. Structured models under graphic form such as bayesian networks should be part of the risk management and dependability major  tools  The WG objective is on one hand to promote capacities of modelling and computing of probabilistic graphic models ( bayesian networks, evidential networks, probabilistic relational networks, bayesian dynamic networks, continuous and hybrid bayesian networks), on another hand to illustrate their implementation on various levels of complexity risk management and dependability problems . Finally this WG aims to bring to the foreground problems which could be seized by industrial and academic communities.



Would you attend this WG, please register at : Philippe WEBER (CRAN) , Christophe SIMON (CRAN)

Exchange space Haut de page
to reach this working group 

This space is reserved to WG members., you must identify to reach it

The meeting
was held on: Tuesday, January 7, 2014