Noise-activated barrier crossing in multiattractor dissipative neural networks

Area of Science:

• Computational neuroscience
• Nonlinear dynamics

Background:

• Investigating noise-activated transitions between coexisting attractors in chaotic spiking networks.
• Understanding attractor hopping and memory conservation at low noise levels.

Purpose of the Study:

• To analyze attractor dynamics and escape probabilities in chaotic spiking networks under varying noise conditions.
• To introduce a metric for evaluating the resilience of biological rhythms.

Main Methods:

• Simulating chaotic spiking networks with varying noise levels.
• Analyzing attractor hopping events and escape probabilities.
• Deriving pseudoactivation energies for limit cycle attractors.

Main Results:

• At low noise, attractor hopping involves discrete bifurcation events preserving initial condition memory.
• At higher noise, attractor lifetime follows detailed balance, with less coherent attractors acting as sinks.
• Escape probability exhibits an activation law, enabling pseudoactivation energy assignment.

Conclusions:

• Pseudoactivation energies provide a useful metric for assessing biological rhythm resilience to perturbations.
• The study elucidates the transition mechanisms between attractors in noisy complex systems.