Spiking neural network review reveals 4 key collective dynamics patterns

A new systematic review titled "Collective dynamics in spiking neural networks" by Afifurrahman and Mohd Hafiz Mohd (arXiv:2410.23876) uses the PRISMA methodology to identify emergent behaviors in excitatory-inhibitory (E-I) spiking neural networks. The authors screened 491 papers from the SCOPUS database (2014–2024), applying strict inclusion/exclusion criteria that narrowed the field to just six relevant studies. Their analysis uncovered four generic collective dynamics regimes: synchrony, irregular spiking, stationary states (where firing rates remain constant), and oscillatory patterns. Notably, the Quadratic Integrate-and-Fire (QIF) neuron model was the most widely used tool across the selected literature, valued for its computational efficiency in large-scale simulations.

The review emphasizes that these collective behaviors arise from the interplay of three factors: intrinsic neuronal characteristics (e.g., ion channel dynamics), network topology (connectivity patterns), and external stimuli (input signals). For professionals in tech, this matters because understanding spiking neural dynamics is foundational for neuromorphic computing and brain-inspired AI hardware. Spiking neural networks (SNNs) promise lower power consumption and temporal processing advantages over traditional deep learning. By mapping the known regimes and the QIF model's robustness, this paper provides a concise entry point for researchers and engineers looking to explore SNN-based systems without wading through decades of computational neuroscience literature.