Question re: fitness graph

[rshepard]

What causes the "banding" effect on the loose self-checking? Is that a glitch in the display, or are there actually values that the circuit never "hits"? Just curious, as I can't recall seeing that before. :confused:

[michaelgarvie]

there actually values that the circuit never "hits"

The fitness function is actually discrete 1/(1+u) where u is undiagnosed faults with some constants to get it in a nice range. The current experiment takes only one evaluation to calculate a score. Most previous runs used several in which their fitness was averaged, hence the intermediate values filling in the bands.