In biological neural systems, activity patterns that can be related to the properties of critical states of statistical mechanics have obtained wide attention, mostly as fingerprints of computational optimality. Commonly, in a given system, a single critical state with well-defined critical exponents that correspond to well-known universality classes of statistical physics is expected to occur. We  first demonstrate that this expectation is unjustified [1]. Moreover, we present first experimental evidence of transitions between different critical regimes during the development of in vitro neuronal cultures, and show how a model based on fundamental biological arguments only, reproduces these transitions, at maintained interaction topology [2].  Finally, we present evidence that small size neural networks of the size of a cortical (micro)column follow in their development process a universal behavior, i.e., their development is essentially independent of the specifics of the neuronal dynamics and of the underlying network topology. These results not only question some more dogmas of the current understanding of neuronal systems, they also offer, more generally, a novel perspective towards understanding the behavior of complex interacting many-body systems - like cortical neural networks - via suitably generalized notions of statistical physics. Our results finally suggest that for understanding biological neural networks, the microscopic focus in the description of biological computational units may be dismissed, for a better-suited mesoscopic approach focusing on interacting networks of higher-level physiological modules.   
[1] K. Kanders, T. Lorimer, and R. Stoop, CHAOS 27, 047408 (2017)
[2] K. Kanders, H. Lee, N. Hong, Y. Nam, and R. Stoop, Nat. Comm. Phys., accepted (2019)

 

海报