Spiking networks and their rate-based equivalents: does it make sense to use Siegert neurons?

Neuronal simulations fall in two broad classes: ones that use spiking neurons and ones that don’t. While spiking models match biology better than rate-based systems, computationally they can be quite expensive. The literature offers some attempts to find and use rate-based neuron models that capture important properties of spiking units. One of the most rigorous approaches [1] approximates the output rate of leaky integrate-and-fire neurons (LIF) for Poisson input trains by analyzing the subthreshold activity of the neuron [2]. This approach, the Siegert neuron, is shown in Fig. 1.