This issue list a series of experiments which needs to be done in order to assess the performance of the various evolutionary algorithms:
Differential Evolution (Compute the average/median fitness for each generation)
(Algorithm, Difficulty, N° of generations, Size of the population, Architecture type, Weights update)
[x] DE, Easy, 100, 100, Shallow, Shallow
[x] DE, Normal, 100, 100, Shallow, Shallow
[x] DE, Hard, 100, 100, Shallow, Shallow
[x] DE, Easy, 100, 100, Shallow, Normal
[x] DE, Normal, 100, 100, Shallow, Normal
[x] DE, Hard, 100, 100, Shallow, Normal
[x] DE, Easy, 100, 100, Deep, Shallow
[x] DE, Normal, 100, 100, Deep, Shallow
[x] DE, Hard, 100, 100, Deep, Shallow
[x] DE, Easy, 100, 100, Deep, Normal
[x] DE, Normal, 100, 100, Deep, Normal
[x] DE, Hard, 100, 100, Deep, Normal
[x] DE, Easy, 100, 100, Wide, Shallow
[x] DE, Normal, 100, 100, Wide, Shallow
[x] DE, Hard, 100, 100, Wide, Shallow
[x] DE, Easy, 100, 100, Wide, Normal
[x] DE, Normal, 100, 100, Wide, Normal
[x] DE, Hard, 100, 100, Wide, Normal
[x] DE, Easy, 100, 100, Wider, Shallow
[x] DE, Normal, 100, 100, Wider, Shallow
[x] DE, Hard, 100, 100, Wider, Shallow
[x] DE, Easy, 100, 100, Wider, Normal
[x] DE, Normal, 100, 100, Wider, Normal
[x] DE, Hard, 100, 100, Wider, Normal
[x] DE, Easy, 300, 100, Shallow, Shallow
[x] DE, Normal, 300, 100, Shallow, Shallow
[x] DE, Hard, 300, 100, Shallow, Shallow
[x] DE, Easy, 300, 100, Shallow, Normal
[x] DE, Normal, 300, 100, Shallow, Normal
[x] DE, Hard, 300, 100, Shallow, Normal
[x] DE, Easy, 300, 100, Deep, Shallow
[x] DE, Normal, 300, 100, Deep, Shallow
[x] DE, Hard, 300, 100, Deep, Shallow
[x] DE, Easy, 300, 100, Deep, Normal
[x] DE, Normal, 300, 100, Deep, Normal
[x] DE, Hard, 300, 100, Deep, Normal
[x] DE, Easy, 300, 100, Wide, Shallow
[x] DE, Normal, 300, 100, Wide, Shallow
[x] DE, Hard, 300, 100, Wide, Shallow
[x] DE, Easy, 300, 100, Wide, Normal
[x] DE, Normal, 300, 100, Wide, Normal
[x] DE, Hard, 300, 100, Wide, Normal
[x] DE, Easy, 300, 100, Wider, Shallow
[x] DE, Normal, 300, 100, Wider, Shallow
[x] DE, Hard, 300, 100, Wider, Shallow
[x] DE, Easy, 300, 100, Wider, Normal
[x] DE, Normal, 300, 100, Wider, Normal
[x] DE, Hard, 300, 100, Wider, Normal
[x] DE, Hard, 500, 200, Deep, Shallow
[x] DE, Hard, 500, 200, Deep, Shallow
[x] DE, Hard, 1000, 200, Wide, Shallow
[x] DE, Hard, 2000, 200, Wide, Shallow
[x] DE, Hard, 5000, 200, Wide, Shallow (best DE)
[x] DE, Hard, 3000, 300, Wider, Shallow (just an attempt)
NEAT (Compute the average/median fitness for each generation)
(Algorithm, Difficulty, N° of generations, Size of the population, layer size, Elitism)
[x] NEAT, Easy, 100, 100, 4 Yes
[x] NEAT, Normal, 100, 100, 4 Yes
[x] NEAT, Hard, 100, 100, 4 Yes
[x] NEAT, Easy, 100, 100, 4 No
[x] NEAT, Normal, 100, 100, 4 No
[x] NEAT, Hard, 100, 100, 4 No
[x] NEAT, Easy, 100, 100, 16, Yes
[x] NEAT, Normal, 100, 100, 16, Yes
[x] NEAT, Hard, 100, 100, 16, Yes
[x] NEAT, Easy, 100, 100, 16, No
[x] NEAT, Normal, 100, 100, 16, No
[x] NEAT, Hard, 100, 100, 16, No
Reinforcement Learning Version
[x] Compute mean value of the fitness of the pre-trained model.
This issue list a series of experiments which needs to be done in order to assess the performance of the various evolutionary algorithms:
Differential Evolution (Compute the average/median fitness for each generation) (Algorithm, Difficulty, N° of generations, Size of the population, Architecture type, Weights update)
NEAT (Compute the average/median fitness for each generation) (Algorithm, Difficulty, N° of generations, Size of the population, layer size, Elitism)
Reinforcement Learning Version