Speaker of Workshop 4
Will talk about: Function follows dynamics: state-dependency of information flow in neural circuits
Demian Battaglia, after a training in statistical physics, developed an interest ininterdisciplinary research obtaining in 2005 his PhD at SISSA, Trieste, under the directionof Riccardo Zecchina on themes at the interface between information theory, combinatorialoptimization and physics of complex networks. Since his first postdoc at the Laboratory forNeurophysics and Physiology, University Paris Descartes, advised by David Hansel andNicolas Brunel, his investigations are devoted to computational neuroscience andneuroinformatics. In 2009, he joined the Theo Geisel's lab at the Max Planck Institute forDynamics and Selforganization in Göttingen, Germany and since May 2010 he serves asprincipal investigator in a project of the local Bernstein Center for ComputationalNeuroscience about the flexible control of functional interactions and information flowbetween neural circuits. Awarded of a Marie Curie Fellowship, starting from September2013, he will integrate the Institute for Systems Neuroscience, at the University of Aix-Marseilles, France.
Brain function require the control of inter-circuit interactions on time-scales faster thansynaptic changes. In particular, strength and direction of causal influences and informationexchange between neural populations (described by the so-called effective connectivity)must be reconfigurable even when the underlying structural connectivity is fixed. Suchinfluences can be quantified analyzing time-series of neural activity with tools like GrangerCausality, delayed Mutual Information or Transfer Entropy. But how can manifold functionalnetworks stem from fixed structures? Considering model systems at different scales, likeneuronal cultures or cortical multi-areal motifs, we show that ``function and informationfollow dynamics'', rather than structure. Different dynamic states of a same structuralnetwork, characterized by different synchronization properties, are indeed associated todifferent directed functional networks, corresponding to alternative information flowpatterns. Here we discuss how suitable generalizations of Transfer Entropy, taking intoaccount switching between collective states of the analyzed circuits, can provide a pictureof causal interactions and information flow in agreement with a ``ground-truth'' descriptionat the dynamical systems level.