Berlin 2015 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
BP: Fachverband Biologische Physik
BP 8: Neurophysics II
BP 8.7: Talk
Monday, March 16, 2015, 16:15–16:30, H 1058
Assessing network states from subsampled activity — •Anna Levina1,2,3, Theo Geisel1,3, and Viola Priesemann1,3 — 1BCCN Göttingen, Germany — 2MPI MIS, Leipzig, Germany — 3MPI DS, Göttingen, Germany
Experimental studies suggest that neural activity self-organizes close to criticality, as various preparations have shown signatures of it. At criticality, the neural network may profit from the optimal information processing associated with a critical state.
The appeal of the criticality hypothesis for the brain lies in its potential to unveil a fundamental principle of collective neural dynamics and offers an opportunity to relate neuronal circuits to well studied physical systems. When testing for criticality in simulated systems the full information about their activity can be used. Data obtained from brain recordings are limited by subsampling, however, since to date it is impossible to assess the activity of every single neuron in the brain.
Here we discuss how subsampling changes avalanche size distributions, and how it is possible to recover information of the actual network state even under subsampling. To this end, we extend methods from statistical physics and analyse scaling laws for subsampled systems. To demonstrate the generality of our novel approach, we evaluate models from different universality classes and support our results by analytical considerations.