New survey classifies spiking neuron models for bio-plausible AI systems

A new survey paper from researchers Leon Parepko, Danila Shulepin, and Albert Nasybullin provides a structured overview of mathematical models for single-entity spiking neurons. Published on arXiv in July 2026 (arXiv:2607.07429), the 4-page work categorizes spiking neuron models—such as Hodgkin-Huxley, Izhikevich, and leaky integrate-and-fire—alongside discrete and continuous analogs that capture biological membrane potential dynamics. The authors classify models based on common features and specific use cases, helping researchers choose the right abstraction level for tasks ranging from computational neuroscience to neuromorphic engineering.

The survey emphasizes both prevalence and innovation, selecting approaches that are widely adopted or offer novel perspectives. By including discrete and continuous analogs, it bridges the gap between purely spiking models and traditional neural network formalisms. This work is particularly relevant for AI researchers exploring energy-efficient, brain-inspired computing and for neuroscientists simulating neural circuits. The paper also points to related resources like the 2023 DCNA conference proceedings, offering a compact reference for practitioners needing to navigate the landscape of biologically plausible neuron models.