Synchronization in STDP-driven memristive neural networks with time-varying topology

Area of Science:

• Computational Neuroscience
• Neural Networks
• Complex Systems

Background:

• Synchronization is crucial for brain function, yet optimal conditions for robust synchronization in neural networks with plasticity remain elusive.
• Spike-timing-dependent plasticity (STDP) and homeostatic structural plasticity (HSP) are key learning rules influencing neural network dynamics.
• Understanding parameter configurations for stable synchronization is vital for neuroscience and neuromorphic engineering.

Purpose of the Study:

• To determine specific parameter configurations for achieving robust and enduring complete synchronization (CS) and phase synchronization (PS).
• To investigate the effects of STDP and HSP on synchronization in time-varying small-world and random neural networks.
• To identify how network parameters like synaptic time delay, average degree, and rewiring probability influence synchronization stability.

Main Methods:

• Simulated time-varying small-world and random neural networks.
• Implemented spike-timing-dependent plasticity (STDP) and homeostatic structural plasticity (HSP) learning rules.
• Analyzed the impact of parameter variations (P, F, synaptic time delay, average degree, rewiring probability) on complete synchronization (CS) and phase synchronization (PS) metrics.

Main Results:

• Decreasing parameter P (enhancing STDP) and increasing parameter F (synapse swapping rate) consistently promote higher and more stable CS and PS in both network types.
• The influence of increasing F on CS and PS stability depends on network topology when other parameters are not optimal.
• Synaptic time delay can induce intermittent CS and PS, independent of the F parameter.

Conclusions:

• Specific configurations of STDP and HSP parameters, particularly P and F, are critical for achieving stable neural synchronization.
• Network topology and parameters like synaptic time delay play significant roles in modulating synchronization patterns.
• Findings offer insights for designing neuromorphic circuits for efficient information processing through neural synchronization.