This title appears in the Scientific Report :
2016
Distribution of pairwise covariances in neuronal networks
Distribution of pairwise covariances in neuronal networks
Massively parallel recordings of spiking activity in cortical circuits show large variability of covariances across pairs of neurons [Ecker et al., Science (2010)]. In contrast to the low average, the wide distribution of covariances and its relation to the structural variability of connections betw...
Saved in:
Personal Name(s): | Dahmen, David (Corresponding author) |
---|---|
Diesmann, Markus / Helias, Moritz | |
Contributing Institute: |
JARA-BRAIN; JARA-BRAIN Computational and Systems Neuroscience; IAS-6 Computational and Systems Neuroscience; INM-6 |
Imprint: |
2016
|
Conference: | DPG spring meeting 2016, Regensburg (Germany), 2016-03-06 - 2016-03-11 |
Document Type: |
Abstract |
Research Program: |
The Human Brain Project Supercomputing and Modelling for the Human Brain Theory, modelling and simulation |
Publikationsportal JuSER |
Massively parallel recordings of spiking activity in cortical circuits show large variability of covariances across pairs of neurons [Ecker et al., Science (2010)]. In contrast to the low average, the wide distribution of covariances and its relation to the structural variability of connections between neurons is still elusive. Here, we derive the formal relation between the statistics of connections and the statistics of integral pairwise covariances in networks of Ornstein-Uhlenbeck processes that capture the fluctuations in leaky integrate-and-fire and binary networks [Grytskyy et al., Front. Comput. Neurosci. (2013)]. Spin-glass mean-field techniques [Sompolinsky and Zippelius, Phys. Rev. B (1982)] applied to a generating function representing the joint probability distribution of network activity [Chow and Buice, J. Math. Neurosci. (2015)] yield expressions that explain the divergence of mean covariances and their width when the coupling in the linear network approaches a critical value. Using these relations, distributions of correlations provide insights into the properties of the structure and the operational regime of the network. |