A Modified Experiment for Fatigue
- Modifying fatigue recovery rate so that it equals fatigue
fixes the problem
- Enough activity persists so that the CA persists
- In a given step, about half of the 80 neurons fire
- The neurons that fire change each time step
- All of the firing neurons are in the CA
- To some degree this now works in the standard CANT model
- This modification is a bit weak.
- We've traditionally set
recovery at half the rate of fatigue so that neurons remain
active a third of the time
- It is likely that a larger net or a slightly modified connectivity
would allow persistance at a 2-1 recovery to fatigue ratio
- Depending on the topology, a number of neurons are needed for a
CA to persist
