Taming the model zoo: a unified view on correlations in recurrent networks

Grytskyy, Dmytro; Helias, Moritz; Tetzlaff, Tom; Diesmann, Markus

doi:10.1186/1471-2202-13-s1-p147

Cited by 2 publications

(1 citation statement)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For binary neurons the mean-field approach takes random fluctuations into account to accurately predict the average activity in such networks [7] and expressions for covariances follow from a master equation [8], both briefly reviewed here for completeness. In our recent work we have shown how to map different network models, including binary networks, onto linear dynamics [9]. Binary neurons with a strong non-linear Heaviside gain function are inaccessible to the classical treatment [8].…”

mentioning

confidence: 99%

Invariance of covariances arises out of noise

Grytskyy¹,

Tetzlaff²,

Diesmann³

et al. 2013

AIP Conference Proceedings

View full text Add to dashboard Cite

Correlated neural activity is a known feature of the brain [1] and evidence increases that it is closely linked to information processing [2]. The temporal shape of covariances has early been related to synaptic interactions and to common input shared by pairs of neurons [3]. Recent theoretical work explains the small magnitude of covariances in inhibition dominated recurrent networks by active decorrelation [4,5,6]. For binary neurons the mean-field approach takes random fluctuations into account to accurately predict the average activity in such networks [7] and expressions for covariances follow from a master equation [8], both briefly reviewed here for completeness. In our recent work we have shown how to map different network models, including binary networks, onto linear dynamics [9]. Binary neurons with a strong non-linear Heaviside gain function are inaccessible to the classical treatment [8]. Here we show how random fluctuations generated by the network effectively linearize the system and implement a self-regulating mechanism, that renders population-averaged covariances independent of the interaction strength and keeps the system away from instability.

show abstract

mentioning

confidence: 99%