1st International ICST Workshop on Tools for solving Structured Markov Chains

Research Article

Discrete Time Stochastic Automata Networks: using structural properties and stochastic bounds to simplify the SAN

Download161 downloads
  • @INPROCEEDINGS{10.4108/smctools.2007.1984,
        author={J.M. Fourneau},
        title={Discrete Time Stochastic Automata Networks: using structural properties and stochastic bounds to simplify the SAN},
        proceedings={1st International ICST Workshop on Tools for solving Structured Markov Chains},
        proceedings_a={SMCTOOLS},
        year={2010},
        month={5},
        keywords={Discrete Time Markov Chain Stochastic Bounds Stochastic Automata Networks Aggregation},
        doi={10.4108/smctools.2007.1984}
    }
    
  • J.M. Fourneau
    Year: 2010
    Discrete Time Stochastic Automata Networks: using structural properties and stochastic bounds to simplify the SAN
    SMCTOOLS
    ICST
    DOI: 10.4108/smctools.2007.1984
J.M. Fourneau1,*
  • 1: INRIA project MESCAL Laboratoire Informatique de Grenoble CNRS UMR 5147 Montobonnot, France
*Contact email: jmf@prism.uvsq.fr

Abstract

The Stochastic Automata Network (SAN in the following) methodology is in general associated to exact numerical analysis taking into account the tensor decomposition of the chain to improve matrix vector product. Here, we advocate a completely different approach: use the automata, their properties and the definition of tensor operations to derive structural properties of the chain and stochastic bounds to simplify the SAN. We focus on lumpability and stochastic bounds. Rather than a complete theory which remains to establish, we present a real example: the computation of the loss rate for a multistage ATM switch.