Section: New Results

Probabilistic Activity Description Language

Participants : Elisabetta de Maria, Sabine Moisan, Jean-Paul Rigault.

Since the arrival of E. De Maria in the STARS team in September 2018, we work on the conception of a probabilistic framework for human behavior representation. The goal is to propose (i) a textual language for the description of activities which takes uncertainty into account; (ii) a formal probabilistic model to represent behaviors. Such a model will be tested and validated using experimental data coming from Alzheimer's patients. We will use temporal data resulting from different sensors and corresponding to patients playing with serious games. This will be the topic of T. L'Yvonnet's PhD starting in December. E. De Maria's main researches concern the investigation of the dynamic behavior of biological neuronal networks, using Leaky Integrate and Fire (LIF) neuronal networks, whose temporal dimension is crucial (the state of each neuron is computed taking into account not only present inputs but also past ones). This year, we used the PRISM language to model LIF neuronal networks as probabilistic reactive systems and we proposed an algorithm which aims at reducing the number of neurons and synaptical connections of these networks [42].