Critical dynamics in associative memory networks.

Area of Science:

• Computational neuroscience
• Complex systems theory
• Neural network dynamics

Background:

• Criticality in neural networks, marked by scale-free avalanche distributions, is linked to optimal information storage.
• Self-regulatory mechanisms are proposed to underlie critical brain dynamics.
• The role of structural connectivity, particularly Hebbian learning, in maintaining criticality is not fully understood.

Purpose of the Study:

• To investigate how Hebbian learning affects the criticality of neural network dynamics.
• To identify mechanisms that can stabilize criticality in the presence of Hebbian learning.
• To explore the implications for information storage capacity and memory flexibility.

Main Methods:

• Simulations of neural networks incorporating Hebbian learning rules.
• Introduction of short-term synaptic dynamics (synaptic depression and facilitation).
• Implementation of homeostatic synaptic weight adaptation.
• Analysis of network dynamics for scale-free avalanche size distributions.

Main Results:

• Hebbian learning alone prevents simultaneous information storage and criticality.
• Short-term synaptic dynamics or homeostatic adaptation can stabilize critical network states.
• Heterogeneous maximal synaptic strengths do not prevent criticality if learning alternates with recovery periods.

Conclusions:

• Hebbian learning requires complementary mechanisms like synaptic plasticity or homeostatic regulation to maintain neural network criticality.
• These findings offer insights into memory flexibility in aging and synaptic plasticity theories.
• Stabilizing critical dynamics is crucial for maximizing information processing in neural systems.