Stochastic model checking by stochastic comparison

old_uid6102
titleStochastic model checking by stochastic comparison
start_date2009/01/23
schedule14h-15h
onlineno
location_infosalle 847
summaryWe propose a new method to verify Markov chains. The verification is performed by transient or steady-state distributions of the considered chain. We use stochastic comparison techniques to obtain bounding measures in order to verify the considered chain. These measures are obtained by the construction of a bounding chain to the initial chain that can be very large. Bounding models have to be easier to analyze that let to consider models for which numerical analysis are difficult or impossible. We have explored some schemes of construction of bounding chains as lumpability and class C matrices. We have also developed other methods of construction of bounding chains for censored Markov chains. Clearly, bounding measures can not always lead to a conclusion. In this case, the bounding model must be refined if it is possible. We have showed that bounding techniques that we propose are appropriate for the verification of Markov chains and they may reduce drastically the verification time.
responsiblesBaerecke