Robust cortical criticality and diverse dynamics resulting from functional specification.

Area of Science:

• Computational neuroscience
• Artificial neural networks

Background:

• The brain's layered structure and criticality are recognized, but their impact on self-organized criticality is under-explored.
• Understanding how network architecture influences critical dynamics is key for neural computation.

Purpose of the Study:

• To investigate how input, output, and computational layer specification affect self-organized criticality in neural networks.
• To explore the role of network structure in adapting to external stimuli.

Main Methods:

• Construction of heterogeneous network structures using a leaky neuron model.
• Analysis of network dynamics under varying degrees of recurrence and feedforward connections.
• Simulation of network responses to different strengths of external stimuli.

Main Results:

• Layer specification leads to robust criticality, largely insensitive to external stimulus strength.
• Low recurrence explains subcriticality, offering an alternative to high-frequency input explanations.
• Networks with sufficient feedforward connections can achieve criticality/supercriticality, unlike fully recurrent networks.

Conclusions:

• Functional and structural specification, along with external stimuli, are critical for network dynamics.
• Leaky neuron networks operating near critical points show promise for artificial neural networks.
• Robust criticality achieved through layer specification offers a mechanism for stimulus adaptation without plasticity.