- The fatigue mechanism we've been using for years decreases to, at most, 0 when a neuron does not fire.
- We tried a modified version which continued to fatigue below zero.
- When fatigue was < -.25, we changed the fatigue increase mechanism.
Now it was
- This improved the fit to the biological data, up to about 91%.
- We tried several different negative fatigue increase mechanisms, and this worked best.
