Vous êtes ici : GIPSA-lab > Animation > Séminaires

Séminaire du département Images et Signal du 31/01/2013 à 15h00


Function follows dynamics: state-dependency of information flow in neural circuits

Intervenant : Demian BATTAGLIA, Max Planck Institute for Dynamics and Self-Organisation, Goettingen (Allemagne)

Lieu : Salle Chartreuse


Résumé :

Brain function require the control of inter-circuit interactions on time-scales faster than synaptic changes. 
In particular, strength and direction of causal influences and information exchange between neural populations (described by the so-called effective or directed functional connectivity) must be reconfigurable even when the underlying structural connectivity is fixed. 
Such influences can be quantified analyzing time-series of neural activity with tools like Granger Causality or its model-free generalization, Transfer Entropy. 
But how can manifold functional networks stem from fixed structures? 

Such a general question is explored through computational studies of systems at different scales, including "in silico" models of in vitro cultures of dissociated neurons or of meso-scale motifs of interacting cortical areas. 
Profiting of the advantages provided by a computational framework, in which the ground-truth structure of the analyzed systems is known and in which their dynamics can be characterized with full precision, we show that ``function and information follow dynamics'', rather than structure. 
Different dynamic states of a same structural network, characterized by different synchronization properties, are indeed associated to different directed functional networks, corresponding to alternative information flow patterns. 
Spontaneous or induced through ad hoc pertubations switching between multiple dynamical states leads therefore to a reconfiguration of network-wide functional interactions without need of parallel structural changes.

Here we discuss thus how suitable generalizations of Transfer Entropy, taking into account switching between collective states of the analyzed circuits, can provide a picture of causal interactions and information flow in agreement with ``real'' descriptions at the dynamical systems level, illustrating also applications to the extraction of connectivity from calcium imaging experiments.

GIPSA-lab, 11 rue des Mathématiques, Grenoble Campus BP46, F-38402 SAINT MARTIN D'HERES CEDEX - 33 (0)4 76 82 71 31