Some ongoing projects

Some ongoing projects are briefly presented here and because science is collaboration, schematic and fun, some of our sketches are provided too.
 

Abstraction of structure and dynamics

Information simplification is a current challenge in brain modeling. Mathematical and simulation technics are developped to simplify both anatomical structure and electrical activities of neural networks. This is done either theoretically (e.g., using general systems) or in collaboration with biologists to achieve plausible abstractions (e.g., simplifying the dendritic arborisation of neurons).



Verification of temporal properties of neural graph archetypes


There are experimental evidences that reccurent patterns of neural connections (with well defined behaviors) exist in the brain. Being able to dispose of an alphabet of these structures (or archetypes) and their behaviors would allow, connecting them, to enhance our understanding of brain structure/behavior interactions. To automatically verify the dynamics (as temporal properties) of these archetypes, they are implemented as reactive systems using synchronous programming languages. This implementation offers the possibility of configuring electronic circuits after the automated verification of their behaviors.


 

Firing delays and synchronization in stochastic graphs

Using stochastic neurons and graphs allows, while keeping a discrete approach, studying large networks of neurons at the average. Then, many questions about the behavior of the network can be explored: How long will it be for the whole neurons to fire together? What structural and behavioral parameters lead to periodic firing activities? Which sub-assemblies of neurons frequently fire together? Etc. However, exploring the behavior of these networks can be hard analytically for some parameters. It is useful then to use in silico experiments to explore automatically these unknown behaviors. Having well designed, reusable and reproducible abstract simulators for such stochastic and parallel models also requires the original definition of new general mathematical (computational) structures while having efficiently implementable programs. This project constitutes a good example of defining a coherent multilevel mathematical and operational framework from modeling to simulation.