Neuron Models

Models and Toothbrushes
The best models are compartmental models. Some neurophysiologists call them explanations.
Of course, they're a bit computationally expensive, so most AI people work with simple point models.
The simplest is integrate and fire (McCulloch and Pitts or Lapicque).
That's been used by Hopfield too.
There's a fair bit of work with Leaky Integrate and Fire.
OReilly and others avoid spiking and have continuous output models.
I also think Boltzmann machines that fire regularly without input are good.
Almost all of my work has been with Fatiguing Leaky Integrate and Fire neurons.
I'm starting to play with compartmental models.
Note that there are lots of different types of neurons.
Izhikevich has a good two parameter model for accounting for different types of neurons.
One big division (oversimplified) is between inhibitory and excitatory neurons.