Leaky Integrate and Fire Neurons

• What happens to the energy in and IF neuron when the neuron doesn't spike.
• There are two obvious choices, lose it all or keep it all.
• If you lose it all, I think, you have the Lapicque version.
• If you keep it all, little bits of energy over lots of steps allows the neuron to spike (I think the McCullouch Pitts version).
• It doesn't seem that real neurons do either of these things.
• A subthreshold current can get neurons to fire quite regularly.
• A really low amount of current, doesn't.
• A widely used model (for biological neurons) is the leaky integrate and fire (LIF) model.
• With a LIF model, if a neuron does not fire, it loses some of its activation.
• The activation leaks away. I used to do this by multiplying activation by a constant between 1 and 0. The equations usually have a leak parameter.
• If the neuron fires, it loses its activation. This is typically set to a reset value. That can then drift to the base value using leak.