Telecom parisTech, salle C49: comment y aller
Abstract: The arrival of new complex processors has made the time analysis of the programs more difficult while there is a growing need to integrate uncertainties from all levels of the embedded systems design. Probabilistic and statistical approaches are one possible solution and they require appropriate proofs in order to be accepted by both scientific community and industry. Such proofs cannot be limited at processor or program level and in this talk we provide hints on the possible interaction between different design levels by using the probabilistic formulation as compositional principle.
For the last two decades, complex
event processing under uncertainty has been widely studied,
but, nowadays, researches are still facing difficult
problems as combinatorial explosion or lack of
expressiveness. Numerous approaches have been proposed like
automate-based methods, probabilistic Petri-net, stochastic
context free grammars, or composed methods using first-order
logic and probabilistic graphical models.
Each technique has its own pros and cons that rely on the problem structure and underlying assumptions. In our case, we want to propose a model providing likelihood of a complex event from long data streams produced by a simple, but large system, in a reasonable amount of time. Furthermore, we want this model to be able to consider prior knowledge on data streams with an high degree of expressiveness.