Is cortical criticality unique?

TL;DR

This paper investigates whether different indicators of criticality in neural networks point to a single critical point or multiple, revealing that avalanche criticality and edge-of-chaos are distinct phenomena.

Contribution

The study demonstrates that avalanche criticality and edge-of-chaos are separate phenomena in neural networks, challenging the assumption of a unique critical point for neural optimization.

Findings

Avalanche criticality does not necessarily coincide with the edge-of-chaos.

Different criticality fingerprints may indicate distinct phenomena.

Neural networks can operate at multiple critical states independently.

Abstract

There are indications that for optimizing neural computation, neural networks - including the brain - operate at criticality. Previous approaches have, however, used diverse fingerprints of criticality, leaving open the question whether they refer to a unique critical point or whether there could be several. Using a recurrent spiking neural network as the model, we demonstrate that avalanche criticality does not necessarily lie at the dynamical edge-of-chaos and that therefore, the different fingerprints indicate distinct phenomena with an as yet unclarified relationship.