MAPS: Mappings, Adaptation, Plasticity and Spatial computation

Prochaine réunion

Février 2011 à Marseille

MAPS is a research project partly funded by the french ANR (Agence Nationale de la Recherche).

Duration: 36 months from 12/2007

Four partners in computer science, neuroscience, experimental psychology.


Computational neurosciences gathers a set of scientific domains which try to identify the computational mechanisms that accompany the functioning of the nervous system. The main idea that federates the "MAPS" project is the notion of spatial computation. Indeed, our project is aimed at re-examining the relationship between structure and function, taking into account the topological (spatial aspects) and hodological (connectivity) constraints of the neuronal substrate. We think that those constraints are fundamental for the understanding of integrative processes, from the perception level to the motor level and the initiation of coordinated actions. This approach will have important consequences for understanding the functional organization of brain networks involved in these tasks.

Based on two different theoretical approaches, respectively related to the architecture and the learning in neural maps, we will propose new principles of neural computation. These principles will be compared to data provided by investigations studying the ascending propagation of visual information along the pathway V2-MT-MST and the lateral propagation which is particularly important when a target is moving. However, the neural maps under study will be not only cortical sensory maps, but also sub-cortical maps, in particular the one found in the deep layers of the superior colliculus which is involved in the control of orienting movements.

In order to improve the analysis of such neural computation in the modeling of the living, this study will be held both at the level of the mean population activity frequency, and at the level of the discrete discharges of individual neurons. More than a constraint, we think that bridging the gap between those two levels of description will be crucial for a better understanding of the spatio-temporal characteristics of the neural computations.

The strength in our proposal is to gather scientists from three distinct domains (computer science, biology and experimental psychology) onto a common research object that will benefit to all three domains. Each of these domains will be involved in every step of the workbench and the scientific complementarity of partners will allow gathering data, expertise and knowledge at the level of computational mechanisms, physiology and behavior. The project is roughly organized along three milestones:

  • Milestone 1 : Neural architectures and the development of informational pathways
  • Milestone 2 : Synaptic plasticity mechanisms and reinforcement-driven task learning.
  • Milestone 3 : testing those computation principles in three domains:
    1. modelling and experimental studies of the coding of saccadic amplitude and adaptive processes
    2. modelling and experimental studies on the mapping between visual and oculo-motor spaces
    3. dynamical system modelling and experimental studies on the dissociation between conscious perception and visuo-motor adjustments

For each one of these three milestones, a first 6-month part will be devoted to the gathering of new experimental and physiological data, the identification of structural and functional constraints and the analysis of existing models in order to estimate the best possible architectures. We will then product a critical review on the choice of the mechanisms to integrate, with the production of a synthetic bibliographical document.

Then, on every modeling milestone, the studies will undergo a progressive inclusion of integration levels, from the feed-forward models to the lateral ones, and then to the feedback. In parallel, several experimental studies on animal physiology and experimental psychology will be conducted, according to a common methodology : model construction, experimentation and confrontation with experimental data. The experimental data and the proposed model will eventually be published in reference journals.

GlossyBlue theme adapted by David Gilbert
Powered by PmWiki