My thesis illustrates a research on cell assemblies, groups of closely
connected, synchronously activating neurons, which are thought to be the
units of memory, and on the techniques of statistical physics and
inference for the study of interacting neurons. A new method to unveil
cell assemblies from neuronal data is illustrated and applied to
multi-electrode recordings in the prefrontal cortex of rats during
performance of a task and during the preceding and following sleep epochs.
The method is based on the inference of an Ising network of effective
interactions between the neurons and on the simulation of the inferred
model in the presence of a global uniform drive: as the drive increases,
configurations of high activity (cell assemblies) are unveiled, which
activate in the data on time scales of tens of ms, in the presence of
transient stimuli. Comparison of the interaction networks and of the
results of the simulations across the three experimental phases reveals
empirical rules for the modification of cell assemblies. The inferred
model is also exploited to estimate the reactivation (replay) of the cell
assemblies during sleep, important for memory consolidation. The aspects
of temporal ordering of the neuronal activity are studied through the
inference and sampling of a generalized linear model.
RECRUITMENT DIRECT ACCESS
The Department of Physics
The ENS Physics Laboratory 