Antonio Galves, University of S.Paulo, Brazil


Context tree models in the brain


Galves and Loecherbach (2013) introduces a new class of stochastic processes describing networks of spiking neurons. These processes are systems of interacting chains with memory of variable length. Each chain in the system describes the spiking activity of a single neutron. Processes in this class are not markovian as the spiking rate of each neuron depends on the history of the system after the last spiking time of that neuron. This class of models extends the class of context tree models introduced in the seminal paper Rissanen (1983). In this talk some recent results about these models will be presented, including a consistent statistical selection procedure to retrieve the graph of interactions characterizing the system.