1.1 What is a spiking neuron?

In this chapter, we will study the most fundamental unit of building SNNs -- the Spiking Neuron. In the previous chapter you were introduced to Biological Neurons and their common types. Here’in, we will learn how to simulate them with various levels of fidelity, using the neuroscience principles underlying the biological neurons.

A section on the functional comparison between spiking neurons and artificial neurons is in preparation and will appear in a later release; it will:

• highlight the inherent temporality and sparsity of spiking neurons, and

• intuitively explain the relation between spiking and artificial neurons

1.1.1Spiking Neurons¶

Spiking Neurons are electro-mathematical abstraction of biological neurons. They are generally not very detailed representations of biological neurons, rather simple enough to reproduce their intended spiking behaviour. The major characteristics of biological neurons -- that are of common interest to mimic (via spiking neurons) are:

• Accounting incoming action-potentials: Spiking neurons simulate this behavior by integrating the incoming action-potentials into their membrane potential/voltage -- either in a decaying or non-decaying fashion (more details later). Assuming the incoming action-potentials positively contribute, the spiking neuron’s potential/voltage increases with time and eventually reaches/crosses a certain set voltage threshold.

• Generating an output action-potential: Spiking neurons simulate the generation of action-potential by producing a binary/graded spike, where a binary spike implies a binary value $\in \{0, 1\}$ and a graded spike implies an integer value $\in \mathbb{Z}^+ > 1$. Note that the values of spikes are also sometimes referred as their amplitude.

• Resetting the membrane potential: Spiking neurons simulate the resetting of membrane potential/voltage via two common methods: hard-reset and soft-reset, where hard-reset implies setting the neuron’s voltage to 0, whereas soft-reset implies setting the neuron’s voltage to a value that is equal to the neuron’s current voltage subtracted by its assumed voltage threshold. The difference between these two will be clear in the later sections.

• Entering into refractory state: Spiking neurons simulate this behavior by keeping their membrane potential/voltage set at 0 (in case of hard-reset, the most usual case) or at a subtracted value (in case of soft-reset) for a certain number of time-steps. This characteristic mimics the neurobiological behaviour of the neurons to discourage producing an action potential immediately after firing one.

• Propagating the action-potential along axon: Spiking neurons generally do not simulate this behaviour, except for the spatial spiking neuron models; whose neural dynamics incorporate this behaviour as a delay (effected in simulation time-steps) in action potential propagation through the modeled axon to the axon-terminals.

Note that we have subtly introduced the concept of point and spatial spiking neuron models here. While researching in SNNs, you will see that a majority of the SNN models are built with point spiking neurons. Our next chapter dives into different kinds of Point Neuron and Spatial Neuron models.

1.1.1.1Topic on Spiking Neurons summary?¶

@incollection{snnbook2026-spiking,
  author    = {Gaurav, Ramashish and Pedersen, Jens Egholm and Bogdan, Petrut},
  title     = {{What is a spiking neuron?}},
  booktitle = {{Practical Spiking Neural Networks}},
  publisher = {Open Neuromorphic},
  year      = {2026},
  edition   = {Version 0.8},
  url       = {https://snnbook.net/topics/1_2_spiking},
}