|
How can a neural network encode the dynamics of movement ? Theoretical considerations| old_uid | 2121 |
|---|
| title | How can a neural network encode the dynamics of movement ? Theoretical considerations |
|---|
| start_date | 2007/01/25 |
|---|
| schedule | 10h30 |
|---|
| online | no |
|---|
| location_info | salle de réunion |
|---|
| summary | Contemporary brain theories pose that the emergence of behavioral and cognitive function is based on the interplay between small-scale and large-scale architectures. In particular, functional differentiation is mostly restricted to the more peripheral primary cortical areas, whereas the ensuing cognitive processes rely more on the higher cortical areas. A necessary building block for such a multi-scale architecture is the flexible but robust encoding of movement dynamics within a local area network, i.e. on the smaller local scales. We propose a novel network mechanisms for dynamical coding, which is based on the symmetry breaking of synaptic weights, and prove that an arbitrary low-dimensional dynamics can be stored. As proof of concept we implement an excitator system in a local area network. Excitators are low-dimensional dynamic systems, which are capable of producing many dynamic properties of discrete and rhythmic movements observed in humans. As such we have shown that a single local area network reliably reproduces human movement dynamics and hence qualifies as a functional element in a multi-scale architecture. |
|---|
| responsibles | Marin |
|---|
| |
|