def run():
# 128x128 thumbnails, 1500x1500 rendered images, 1100x810 viewer, grayscale images.
pb = PictureBreeder(128, 128, 1500, 1500, 1100, 810, 'gray', 4)
cfg = config.Config('config')
# Make sure the network has the expected number of outputs.
if pb.scheme == 'color':
cfg.output_nodes = 3
else:
cfg.output_nodes = 1
cfg.pop_size = pb.num_cols * pb.num_rows
pop = population.Population(cfg)
while 1:
pop.run(pb.eval_fitness, 1)
def run():
cfg = config.Config('novelty_config')
cfg.pop_size = 100
cfg.max_fitness_threshold = 1e38
ne = NoveltyEvaluator(4, 'gray')
if ne.scheme == 'color':
cfg.output_nodes = 3
else:
cfg.output_nodes = 1
pop = population.Population(cfg)
while 1:
pop.run(ne.evaluate, 1)
winner = pop.statistics.best_genome()
if ne.scheme == 'gray':
image = eval_gray_image(winner, full_scale * width,
full_scale * height)
elif ne.scheme == 'color':
image = eval_color_image(winner, full_scale * width,
full_scale * height)
elif ne.scheme == 'mono':
image = eval_mono_image(winner, full_scale * width,
full_scale * height)
else:
raise Exception('Unexpected scheme: {0!r}'.format(ne.scheme))
im = np.clip(np.array(image), 0, 255).astype(np.uint8)
im = ne.image_from_array(im)
im.save('winning-novelty-{0:06d}.png'.format(pop.generation))
if ne.scheme == 'gray':
image = eval_gray_image(winner, width, height)
elif ne.scheme == 'color':
image = eval_color_image(winner, width, height)
elif ne.scheme == 'mono':
image = eval_mono_image(winner, width, height)
else:
raise Exception('Unexpected scheme: {0!r}'.format(ne.scheme))
float_image = np.array(image, dtype=np.float32) / 255.0
ne.archive.append(float_image)
def main():
global game_sequence, min_weight, max_weight
game_path = os.path.join(os.path.join(os.getcwd(), "Games"), "FlappyBird")
# genome_path = os.path.join(game_path, "genomes")
config_path = os.path.join(game_path, 'flappy_config')
pop_config = config.Config(config_path)
pop_config.save_best = True
min_weight, max_weight = pop_config.min_weight, pop_config.max_weight
try:
init_population = []
pop_size = pop_config.pop_size
genome_files = sorted(
[f for f in os.listdir() if f.startswith("best_genome_")],
key=lambda x: int(x.split("_")[-1]))
for filename in genome_files[-pop_size:]: # get best ones
with open(os.path.join(os.getcwd(), filename), "rb") as f:
genome = pickle.load(f)
genome.ID = str(len(init_population))
init_population.append(genome)
best_genome = init_population[-1]
for i in range(
len(init_population), pop_size
): # Fill rest with best genome if there is not 50 genomes
new_genome = copy.deepcopy(best_genome)
new_genome.ID = str(i)
init_population.append(new_genome)
gen_number = int(genome_files[-1].split("_")
[-1]) # in order not to override previous generations
except IndexError:
init_population = None
gen_number = -1
pop = population.Population(pop_config, init_population)
pop.generation = gen_number
game_sequence = Sequence(speed, FlappyBirdAPI(True), None,
"flappybirdneat")
pop.run(flappy_fitness, 200)
from ReverseEncodingTree.utils.operator import Operator
has_environment = False
for environment_space in gym.envs.registry.all():
if "LunarLander-v2" in environment_space.id:
has_environment = True
break
if not has_environment:
raise Exception("no environment named LunarLander-v2.")
environment = gym.make("LunarLander-v2")
if __name__ == '__main__':
task_config = config.Config(genome.DefaultGenome, reproduction.DefaultReproduction,
species.DefaultSpeciesSet, stagnation.DefaultStagnation,
"../configures/example/lunar-lander-v2")
# load evolution process.
fitter = FitDevice(FitProcess())
fitter.set_environment(environment=environment, episode_steps=300, episode_generation=5,
input_type=TypeCorrect.List, output_type=TypeCorrect.Value)
# initialize the NeuroEvolution
operator = Operator(config=task_config, fitter=fitter,
node_names={-1: '1', -2: '2', -3: '3', -4: '4',
-5: '5', -6: '6', -7: '7', -8: '8',
0: 'fire engine'},
max_generation=1000, output_path="../output/")
fit = 0
tot = 0.0
net = nn.create_recurrent_phenotype(g)
with open("match_prepared.dat", 'r') as f, open("match_out_prepared.dat", 'r') as f2:
for line_tr, line_te in itertools.izip(f, f2):
if tot > 14000:
break
res = int(net.activate([float(num) for num in line_tr.strip().split()])[0])
if res == int(line_te.strip()):
fit += 1
tot += 1
g.fitness = fit/tot
print "Correct guesses {0}, in total of {1} matches.\n".format(str(fit), str(tot))
conf = config.Config('flappy_config')
conf.save_best = True
try:
init_population = []
pop_size = conf.pop_size
genome_files = sorted([f for f in os.listdir('.') if f.startswith("best_genome_")],
key=lambda x: int(x.split("_")[-1]))
for filename in genome_files[-pop_size:]: # get best ones
with open(os.path.join(os.getcwd(), filename), "rb") as f:
genome = pickle.load(f)
genome.ID = str(len(init_population))
init_population.append(genome)
best_genome = init_population[-1]
for i in range(len(init_population), pop_size): # Fill rest with best genome if there is not 50 genomes
has_environment = False
for environment_space in gym.envs.registry.all():
if "CartPole-v0" in environment_space.id:
has_environment = True
break
if not has_environment:
raise Exception("no environment named CartPole-v0.")
environment = gym.make("CartPole-v0").unwrapped
if __name__ == '__main__':
# load configuration.
task_config = config.Config(genome.DefaultGenome,
reproduction.DefaultReproduction,
species.DefaultSpeciesSet,
stagnation.DefaultStagnation,
"../configures/example/cart-pole-v0")
# load evolution process.
fitter = FitDevice(FitProcess())
fitter.set_environment(environment=environment,
episode_steps=300,
episode_generation=10,
input_type=TypeCorrect.List,
output_type=TypeCorrect.Value)
# initialize the NeuroEvolution
operator = Operator(config=task_config,
fitter=fitter,
node_names={
def __init__(self,
method_type,
logic_type,
max_generation,
display_results=False,
checkpoint=-1,
stdout=False):
"""
initialize the logical task.
:param method_type: the evolution strategy, FS-NEAT, Bi-NEAT, or GS-NEAT.
:param logic_type: the task type, IMPLY, NAND, NOR, or XOR.
:param max_generation: maximum generation of the evolution strategy,
if the generation exceeds the maximum, it will be terminated.
:param display_results: whether result visualization is required.
:param checkpoint: check the statistics point.
:param stdout: Whether outputting the genome information in the process is required.
"""
data_inputs = None
data_outputs = None
if logic_type == LogicType.NAND:
data_inputs = [(0.0, 0.0), (0.0, 1.0), (1.0, 0.0), (1.0, 1.0)]
data_outputs = [(1.0, ), (1.0, ), (1.0, ), (0.0, )]
self.filename = "nand."
elif logic_type == LogicType.NOR:
data_inputs = [(0.0, 0.0), (0.0, 1.0), (1.0, 0.0), (1.0, 1.0)]
data_outputs = [(0.0, ), (0.0, ), (0.0, ), (1.0, )]
self.filename = "nor."
elif logic_type == LogicType.IMPLY:
data_inputs = [(0.0, 0.0), (0.0, 1.0), (1.0, 0.0), (1.0, 1.0)]
data_outputs = [(1.0, ), (1.0, ), (0.0, ), (1.0, )]
self.filename = "imply."
elif logic_type == LogicType.XOR:
data_inputs = [(0.0, 0.0), (0.0, 1.0), (1.0, 0.0), (1.0, 1.0)]
data_outputs = [(0.0, ), (1.0, ), (1.0, ), (0.0, )]
self.filename = "xor."
# load evolution process.
fitter = FitDevice(
FitProcess(init_fitness=4, eval_type=EvalType.ManhattanDistance))
fitter.set_dataset({"i": data_inputs, "o": data_outputs})
# load configuration.
task_config = None
if method_type == MethodType.N:
task_config = config.Config(genome.DefaultGenome,
reproduction.DefaultReproduction,
species.DefaultSpeciesSet,
stagnation.DefaultStagnation,
"../configures/task/logic.n")
self.filename += "fs"
elif method_type == MethodType.FS:
task_config = config.Config(genome.DefaultGenome,
reproduction.DefaultReproduction,
species.DefaultSpeciesSet,
stagnation.DefaultStagnation,
"../configures/task/logic.fs")
self.filename += "fs"
elif method_type == MethodType.BI:
task_config = config.Config(autogenome.GlobalGenome,
bi.Reproduction,
autospecies.StrongSpeciesSet,
stagnation.DefaultStagnation,
"../configures/task/logic.bi")
self.filename += "bi"
elif method_type == MethodType.GS:
task_config = config.Config(autogenome.GlobalGenome,
gs.Reproduction,
autospecies.StrongSpeciesSet,
stagnation.DefaultStagnation,
"../configures/task/logic.gs")
self.filename += "gs"
# initialize the NeuroEvolution
self.operator = Operator(config=task_config,
fitter=fitter,
node_names={
-1: 'A',
-2: 'B',
0: 'A operate B'
},
max_generation=max_generation,
checkpoint_value=checkpoint,
stdout=stdout,
output_path="../output/")
# set whether display results
self.display_results = display_results
self.max_generation = max_generation
def __init__(self,
method_type,
game_type,
episode_steps,
episode_generation,
max_generation,
attacker=None,
noise_level=-1,
display_results=False,
checkpoint=-1,
stdout=False):
"""
initialize the game task.
:param method_type: the evolution strategy, FS-NEAT, Bi-NEAT, or GS-NEAT.
:param game_type: the task type, CartPole_v0 or LunarLander_v2.
:param episode_steps: step parameter of game task
:param episode_generation: generation parameter of game task.
:param max_generation: maximum generation of the evolution strategy,
if the generation exceeds the maximum, it will be terminated.
:param attacker: noise attacker, see evolution/bean/attacker.py.
:param noise_level: noise level.
:param display_results: whether result visualization is required.
:param checkpoint: check the statistics point.
:param stdout: Whether outputting the genome information in the process is required.
"""
game_environment = None
if game_type == GameType.CartPole_v0:
game_environment = gym.make("CartPole-v0").unwrapped
self.filename = "cart-pole-v0."
self.node_name = {
-1: 'In0',
-2: 'In1',
-3: 'In3',
-4: 'In4',
0: 'act1',
1: 'act2'
}
elif game_type == GameType.LunarLander_v2:
game_environment = gym.make("LunarLander-v2")
self.filename = "lunar-lander-v2."
self.node_name = {
-1: '1',
-2: '2',
-3: '3',
-4: '4',
-5: '5',
-6: '6',
-7: '7',
-8: '8',
0: 'fire engine'
}
fitter = FitDevice(FitProcess())
fitter.set_environment(environment=game_environment,
input_type=TypeCorrect.List,
output_type=TypeCorrect.Value,
episode_steps=episode_steps,
episode_generation=episode_generation,
attacker=attacker,
noise_level=noise_level)
# load configuration.
task_config = None
if method_type == MethodType.N:
self.filename += "n"
task_config = config.Config(genome.DefaultGenome,
reproduction.DefaultReproduction,
species.DefaultSpeciesSet,
stagnation.DefaultStagnation,
"../configures/task/" + self.filename)
elif method_type == MethodType.FS:
self.filename += "fs"
task_config = config.Config(genome.DefaultGenome,
reproduction.DefaultReproduction,
species.DefaultSpeciesSet,
stagnation.DefaultStagnation,
"../configures/task/" + self.filename)
elif method_type == MethodType.BI:
self.filename += "bi"
task_config = config.Config(autogenome.GlobalGenome,
bi.Reproduction,
autospecies.StrongSpeciesSet,
stagnation.DefaultStagnation,
"../configures/task/" + self.filename)
elif method_type == MethodType.GS:
self.filename += "gs"
task_config = config.Config(autogenome.GlobalGenome,
gs.Reproduction,
autospecies.StrongSpeciesSet,
stagnation.DefaultStagnation,
"../configures/task/" + self.filename)
# initialize the NeuroEvolution
self.operator = Operator(config=task_config,
fitter=fitter,
node_names=self.node_name,
max_generation=max_generation,
checkpoint_value=checkpoint,
stdout=stdout,
output_path="../output/")
# set whether display results
self.display_results = display_results
self.max_generation = max_generation