We revisit the behavior of a prototypical model of balanced activity in networks of spiking neurons. A detailed study of an appropriate thermodynamic limit for fixed density of connections shows that, when inhibition prevails, the asymptotic regime is characterized by a self-sustained irregular, macroscopic (collective) dynamics rather than being asynchronous. This holds true even for very small coupling strengths. A relationship with the collective chaos observed in standard (unbalanced) heterogeneous networks is also put forward and the role played by clustered states discussed.
Bibliographical noteThe authors acknowledge N. Brunel, F. Farkhooi,
G. Mato, M. Timme, and M. di Volo for enlightening discussions.
This work has been partially supported by the
French government under AMIDEX Grant No. ANR-11-
IDEX-0001-02 (A.T.) and mainly realized at the Max Planck
Institute for the Physics of Complex Systems (Dresden, Germany)
within the activity of the Advanced Study Group
2016/17 “From Microscopic to Collective Dynamics in Neural