Keynotes

Abstract:

When constructing a mathematical model for a given system under study, decisions about characteristics and levels of detail of the model have to be taken. Which choices are appropriate depend on the questions, one wants to answer. It should also depend on available data, such that the model can exploit the information that can be extracted and not suffer too much by what cannot. I will present some examples where a simple model extracted from more biophysical based models can answer specific questions of interest, as long as the simple model is interpreted and used in a suitable way.

Title: Space:the final frontier spatiotemporal dynamics in neural fields

Abstract:

With the ability of experimentalists to record and stimulate with optical techniques, it is possible to determine the behavior of large numbers of neurons as well as to perturb them. In this talk, I consider the behavior of systems of integro-differential equations that arise in the modeling of local cortical areas. First I discuss some issues with the analysis and interpretation of the spatio-temporal data. Next I introduce the class of models and discuss computational and mathematical challenges associated with these nonlocal highly nonlinear systems. I will show several examples of models of the experimental phenomena and thenI will describe recent work on multi-layer models as well as some interesting dynamics that is hidden in simplified discontinuous approximations.

Title :  The visual brain : computing through multiscale complexity

Abstract :

The early visual system is responsible for most of our conscious perception of the visual world. Despite more than 50 years of progresses in Neuroscience, we remain largely ignorant of the computational principles underlying its functions. The main reason for this is the absence of a solid theoretical framework linking multiple biological observations realized at different scales of integration, from the biophysical activation of conductances to the emergence of unified psychological laws (Gestalt), guiding our everyday perception.

Primary visual cortex (V1) in the mammalian brain computes on-the-fly perceptual primitives (form, motion, visual flow) from the feedforward bombardment of retinal events channeled through the thalamus. At the same time, it integrates lateral diffusion within V1 itself and the distributed feedback of higher cortical areas involved in more elaborate cognitive functions. The reverberating activity evoked by the interplay between these three streams has been hypothesized to form the trace of the low-level computational operations written on the “high resolution buffer” of primary cortical areas.

The focus of this talk is to show, from the reading of synaptic echoes recorded in a single V1 neuron, to what extent emerging macroscopic features in low-level perception (Gestalt and motion illusory percepts) can be predicted in primary sensory areas from microscopic levels of neural integration. Various concepts borrowed to non-linear system theory, ultrasonic echography and dynamic systems are presented to account for the functional complexity of a strange matter, the living brain.

Supported by CNRS, ANR, Idex (NeuroSaclay & ICode) and the FET Program (BrainScaleS and “The Human Brain Project”).