Spontaneous Firing from Hypo-Fatigue
- This is pretty much the model we've been using for 15 years.
- We improved the fit to spiking data by modifying fatigue.
- Fatigue always used to have a minimum of zero, but we changed it
so that it could become negative.
- When fatigue got negative enough, the neuron would naturally fire.
- Fatigue was divided by two when it fired and was less than -.25.
- In the spiking data this never helped, but it made things fire
earlier after low input.
- We used to use spontaneous activation, but each neuron just was
randomly selected. Now it's principled.
