In his book Emergence, John Holland descibes two time-based characteristics of neurons with the graphs above.
The time-varying threshold graph on the left indicates the threshold decrease as a function of time since the neuron last fired, with duration (a) being the absolute refractory period. This is a period where after firing, the neuron is incapable of being re-fired.
Neurons exhibit fatigue as the firing rate increases above average. The opposite is also true. A neuron can become more prepared to fire than average as the firing rate drops below the mean value as shown in the graph on the right.
The combination of time-varying thresholds and fatigue provides a mechanism of negative feedback, preventing the system from 'running away' which is the general tendency where the population is greater than two. The first simulation below distributes three 'neurons' in close proximity to illustrate these effects. Stimuli are introduced by clicking on a neuron.
The simulation will generally settle after the intial activity, with a tendency towards pulses of activity as the 'neurons' fatigue, recover, and vary their thresholds of required stimulation. These pulses of activity may take some time before they are observed. The propogation distances have been capped for clarity. Interestingly, the capping of the propogation distances changes the system from being fully-connected, to an iterated system where clusters of 'neurons' can act independantly of others within the field. The attributes of both systems are fully described in W. Ross Ashby's 'Design For A Brain' .
The simulation above, on the left, is the multiple model from the first iteration to allow comparison with the second iteration model on the right. Again the pulses of activity in the right hand model may take some time to appear.
Refreshing the page will generate new distributions.
Investigations will now move towards physical prototypes.