Max Planck Institute for Dynamics and Self-Organization -- Department for Nonlinear Dynamics and Network Dynamics Group
Personal tools
Log in

BCCN Seminar

Wednesday, 06.07.2005 10 s.t.

What's a good Neuron Model?

by Dr. Björn Naundorf
from Max-Planck-Institute für Dynamik und Selbstorganisation, Göttingen

Contact person: Tobias Frederic Niemann


Seminarraum Haus 2, 4. Stock (Bunsenstr.)


With the advent of increasing computer power and better functional characterization of cortical circuitry, it has become possible to simulate large-scale neural networks, taking into account realistic network connectivities. Little attention, however, has so far been paid to the realistic modeling of the spiking dynamics of such networks and, thus, there is an urgent need to identify and characterize simplified neural models which realistically reflect the dynamics of cortical neurons. In my talk, I will first discuss conventional approaches for the construction of simplified neuron models which aim at reproducing membrane potential traces of cortical neurons in in-vitro recordings, or at reproducing the stationary response properties in an in-vivo like regime. It turns out, however, that models that are constructed according to these constraints typically lack a very crucial feature: A realistic dynamical response behavior. I will therefore propose two different approaches which aim at constructing models exhibiting realistic dynamical response properties based on the analysis of the subthreshold dynamics and dynamical action potential initiation of cortical neurons in vitro and in vivo. The identification of models based on these approaches will be an important prerequisite for the realistic modeling of large-scale neural networks and the uncovering of emergent neural computations.

back to overview