Self-organized criticality in neural network models

TL;DR

This paper reviews theoretical models of neural network criticality, focusing on an improved self-organized model that naturally exhibits avalanche statistics aligning with experimental observations.

Contribution

It demonstrates that an enhanced self-organized critical neural network model can replicate experimental avalanche data without parameter tuning.

Findings

Model exhibits critical avalanche behavior matching experimental data

No parameter tuning needed for criticality in the model

Improved model overcomes limitations of earlier spin-based models

Abstract

It has long been argued that neural networks have to establish and maintain a certain intermediate level of activity in order to keep away from the regimes of chaos and silence. Strong evidence for criticality has been observed in terms of spatio-temporal activity avalanches first in cultures of rat cortex by Beggs and Plenz (2003) and subsequently in many more experimental setups. These findings sparked intense research on theoretical models for criticality and avalanche dynamics in neural networks, where usually some dynamical order parameter is fed back onto the network topology by adapting the synaptic couplings. We here give an overview of existing theoretical models of dynamical networks. While most models emphasize biological and neurophysiological detail, our path here is different: we pick up the thread of an early self-organized critical neural network model by Bornholdt and Roehl (2001) and test its applicability in the light of experimental data. Keeping the simplicity of early models, and at the same time lifting the drawback of a spin formulation with respect to the biological system, we here study an improved model (Rybarsch and Bornholdt, 2012b) and show that it adapts to criticality exhibiting avalanche statistics that compare well with experimental data without the need for parameter tuning.