Background process
- Population size = 1000 agents
training set = 25 agents:
| {
|
15/1 fittest agents from foreground
10/24 fittest agents from background
|
- Iteration:
- All 1000 play against each 25 in training set
- Fitness evaluated
- Best half mate and generate new bottom half
- Best 10/24 (with fitness sharing) go to training set
Every foreground iteration:
- Population is reset / seeded with best 100
- New 15/1 best are/is received