|
Résumé |
In large neuronal networks, functions emerge through the collective behavior of
many interconnected neurons. Recent technical development of whole brain imaging
in Caenorhabditis elegans - a nematode with 302 neurons, allowed us to ask if such
emergence reaches down to even the smallest brains. In the first part of this
talk, I will discuss how we use the maximum entropy principle to construct
pairwise probabilistic models for the collective activity of 50+ neurons in C.
elegans. These models successfully predict higher order statistical structure in
the data, the topological features of the structural connectome, and show
signatures of collective behavior. In the second part, I will present two ways of
how perturbing the inferred model of neuronal activity can shed light on the
control principles in the brain, which in turn facilitates future perturbation
experiments. Firstly, by ablating and clamping neurons, we discover that the worm
brain is both robust against damages and efficient in transmitting information.
Secondly, by examining the local information geometry of the model, we find that a
few, "pivotal" neurons account for most of the system's sensitivity, suggesting a
sparse mechanism for control of the collective behavior. Finally, if time allows,
I will briefly describe my current work at ENS on inferring statistical models
with long memory kernel for collective dynamics in a group of social animals. |
