Brain and Mind Institute Seminar
19.03.09 Thursday, 14h15, AAB032
Moritz Helias Bernstein Center for Computational Neuroscience (BCCN),
Albert-Ludwig University Freiburg, Germany
http://www.bccn.uni-freiburg.de/People/members/helias
Structural plasticity controlled by calcium based correlation detection
Abstract:
The connectivity structure of cortical networks was recently found to
exhibit plastic changes even in adult animals [1]: There is a constant
considerable synapse turnover.
It is still unclear, how networks can maintain their functionality
despite this plasticity and what are the mechanisms to ensure network
stability. Prior to addressing these questions, a model of synapse
creation and synapse death has to be developed. The theory of Hebbian
learning assumes the correlation of pre- and postsynaptic spiking
activity to play a crucial role and synapse turnover was indeed found
to be activity dependent [2].
Here we follow the idea of 'correlation based sorting' and present a
postsynaptic selection rule for synapse elimination. Our model
utilizes biologically realistic pathways and constitutes a robust
mechanism of a correlation dependent pruning process. In recurrent
networks it acts as a homeostasis mechanism that guarantees network
stability in certain ranges of activity. In addition, cooperation and
competition between synapses emerges directly from the microscopic
model. This enables neurons to learn patterns of coactive inputs.
In parallel to direct numerical simulations, we derive analytic
results that allow to address the problem of structural plasticity in
recurrent networks on a more abstract level.
back |