In this thesis, theoretical models with increasing levels of abstraction are developed to address questions arising from neuroscience experiments. They are studied using numerical and analytical approaches. With Laurent Venance’s laboratory (Paris), we have developed an ITDP (input-timing-dependent plasticity) protocol model for the plasticity of cortico- and thalamo-striatal synapses. The model has been calibrated with ex vivo data and will be used to determine the presence of synaptic plasticity in vivo, in behavioral experiments aimed at determining the role of cortical and thalamic inputs in motor learning.

At the level of neuronal populations, I have studied the modulation of neuronal collective behaviors by astrocytes, in particular Up-Down synchronization, a spontaneous alternation between periods of high collective activity and periods of silence. I have proposed rate and spiking neural network models of interconnected populations of neurons and astrocytes. They offer explanations of how astrocytes induce Up-Down transitions. Astrocytes are also probably involved in the generation of epileptic seizures, during which neuronal synchronization is impaired. Based on the above models, I have developed a neuron-astrocyte network with a cluster connectivity, showing the transition between Up-Down dynamics and events of very high activity mimicking an epileptic seizure.

Finally, at the level of the brain itself, I studied the standard theory of consolidation, according to which short-term memory in the hippocampus enables the consolidation of long-term memory in the neocortex. I have sought to explain this phenomenon by integrating biological hypotheses – the size of the neocortex explaining the slowness of learning, and neurogenesis in the hippocampus explaining the erasure of its memory – into a model of interconnected neural fields that well reproduces the main features of the theory.