MPIDS Colloquium: Indications for optimality of motor cortex

MPIDS Colloquium

  • Datum: 14.02.2018
  • Uhrzeit: 14:15 - 15:15
  • Vortragende(r): Prof. Moritz Helias
  • Research Center Jülich, Germany
  • Ort: Max-Planck-Institut für Dynamik und Selbstorganisation (MPIDS)
  • Raum: Prandtl Lecture Hall
  • Gastgeber: MPIDS
  • Kontakt: viola.priesemann@ds.mpg.de
The brain processes time-varying input, but it is not known how it achieves its high computational performance. Indeed, neuronal networks can show a rich set of dynamical states. Such states can differ in the measureable spatio-temporal patterns of activity [1]. But they may also differ internally, in terms of their degree to which they produce chaotic activity [2].
We here show that recordings from motor cortex support an operation close to a transition to chaos [3]. We identify this computationally beneficial regime by combining finite-size mean-field theory with massively parallel spike recordings. The theoretical predictions resolve the puzzle how a bal-anced state can be compatible with widely distributed correlations and long-time dynamics.
We then investigate how such networks process input by quantifying how time-varying input sup-presses chaos so that a novel dynamical regime emerges [2]: memory capacity in this regime is op-timal.
Together these findings provide hints towards the operation principles of cortical networks.

Partially supported by the Helmholtz association, Helmholtz young investigator's group VH-NG 1028, RWTH ERS seed fund and the Hans Hermann Voss Stiftung.

[1] Johanna Senk, Karolína Korvasová, Jannis Schuecker, Espen Hagen, Tom Tetzlaff, Markus Diesmann, Moritz He-lias (2018), “Conditions for traveling waves in spiking neural networks”, arXiv:1801.06046 [q-bio.NC]
[2] Jannis Schuecker, Sven Goedeke, Moritz Helias (2016), “Optimal sequence memory in driven random networks”, arXiv:1603.01880 [q-bio.NC]
[3] David Dahmen, Markus Diesmann, Moritz Helias (2016), “Distributions of covariances as a window into the opera-tional regime of neuronal networks”, arXiv:1605.04153 [cond-mat.dis-nn]
Zur Redakteursansicht