def draw_net(net, filename, node_names):
# Create PDF using PUREPLES function
draw_net(net, filename=filename, node_names=node_names)
# Delete text
os.remove(filename)
def run(config_file):
# Load configuration.
config = neat.Config(neat.DefaultGenome, neat.DefaultReproduction,
neat.DefaultSpeciesSet, neat.DefaultStagnation,
config_file)
# Create the population, which is the top-level object for a NEAT run.
# p = neat.Population(config)
p = neat.Checkpointer.restore_checkpoint('neat-checkpoint-4')
# Add a stdout reporter to show progress in the terminal.
p.add_reporter(neat.StdOutReporter(True))
stats = neat.StatisticsReporter()
p.add_reporter(stats)
p.add_reporter(neat.Checkpointer(1000))
#p = neat.Checkpointer.restore_checkpoint('neat-checkpoint-299')
# Run for up to 300 generations.
pe = neat.ParallelEvaluator(multiprocessing.cpu_count(), eval_genome)
winner = p.run(pe.evaluate, 300)
# Display the winning genome.
print('\nBest genome:\n{!s}'.format(winner))
# Show output of the most fit genome against training data.
print('\nOutput:')
cppn = neat.nn.RecurrentNetwork.create(winner, config)
network = ESNetwork(sub, cppn, params)
winner_net = network.create_phenotype_network(
filename='es_hyperneat_winner.png')
input("Winner is found")
for i in range(10):
train(winner_net, network, True)
draw_net(cppn, filename="es_hyperneat")
def _draw_net(net, filename, node_names):
# Create PDF
draw_net(net, filename=filename, node_names=node_names)
# Delete text
os.remove(filename)
def run(config_file):
# Load configuration.
config = neat.config.Config(neat.genome.DefaultGenome,
neat.reproduction.DefaultReproduction,
neat.species.DefaultSpeciesSet,
neat.stagnation.DefaultStagnation, config_file)
# Create the population, which is the top-level object for a NEAT run.
#p = neat.Checkpointer.restore_checkpoint('neat-checkpoint-15')
# Add a stdout reporter to show progress in the terminal.
pop = neat.population.Population(config)
stats = neat.statistics.StatisticsReporter()
pop.add_reporter(stats)
pop.add_reporter(neat.reporting.StdOutReporter(True))
pop.add_reporter(neat.Checkpointer(999))
#p = neat.Checkpointer.restore_checkpoint('neat-checkpoint-299')
# Run for up to 300 generations.
winner = pop.run(eval_multipleGenomes, 300)
# Display the winning genome.
print('\nBest genome:\n{!s}'.format(winner))
# Show output of the most fit genome against training data.
print('\nOutput:')
cppn = neat.nn.FeedForwardNetwork.create(winner, config)
network = ESNetwork(sub, cppn, params)
winner_net = network.create_phenotype_network(
filename='es_hyperneat_winner.png')
input("Winner is found")
for i in range(10):
train(winner_net, network, True)
draw_net(cppn, filename="es_hyperneat")
def run(config_file):
# Load configuration.
config = neat.Config(neat.DefaultGenome, neat.DefaultReproduction,
neat.DefaultSpeciesSet, neat.DefaultStagnation,
config_file)
# Create the population, which is the top-level object for a NEAT run.
p = neat.Population(config)
#
# p = neat.Checkpointer.restore_checkpoint('neat-checkpoint-100')
#
#
# thenet = neat.nn.FeedForwardNetwork.create(p.best_genome, config)
# output = thenet.activate(i)
#
# Add a stdout reporter to show progress in the terminal.
p.add_reporter(neat.StdOutReporter(True))
stats = neat.StatisticsReporter()
p.add_reporter(stats)
p.add_reporter(neat.Checkpointer(0))
# Run for up to 300 generations.
winner = p.run(eval_genomes, 150)
# import gzip
#
# try:
# import pickle as pickle # pylint: disable=import-error
# except ImportError:
# import pickle # pylint: disable=import-error
#
# def save_object(obj, filename):
# with gzip.open(filename, 'w', compresslevel=5) as f:
# pickle.dump(obj, f, protocol=pickle.HIGHEST_PROTOCOL)
#
# def load_object(filename):
# with gzip.open(filename) as f:
# obj = pickle.load(f)
# return obj
# save_object(winner,"neat.pkl")
# winner = load_object("neat.pkl")
# Display the winning genome.
# print('\nBest genome:\n{!s}'.format(winner))
# Show output of the most fit genome against training data.
print('\nOutput:')
winner_net = neat.nn.RecurrentNetwork.create(winner, config)
draw_net(winner_net, filename="neat_torcs_winner_steering")
with open('neat_torcs_steering.pkl', 'wb') as output:
pickle.dump(winner_net, output, pickle.HIGHEST_PROTOCOL)
# Config for FeedForwardNetwork.
config = neat.config.Config(neat.genome.DefaultGenome,
neat.reproduction.DefaultReproduction,
neat.species.DefaultSpeciesSet,
neat.stagnation.DefaultStagnation,
'config_neat_pole_balancing')
# Use the gym_runner to run this experiment using NEAT.
def run(gens, env):
winner, stats = run_neat(gens, env, 500, config)
print("neat_pole_balancing done")
return winner, stats
# If run as script.
if __name__ == '__main__':
# Setup logger and environment.
logger = logging.getLogger()
logger.setLevel(logging.INFO)
env = gym.make("CartPole-v1")
# Run!
winner = run(500, env)[0]
# Save net if wished reused and draw it + winner to file.
winner_net = neat.nn.FeedForwardNetwork.create(winner, config)
draw_net(winner_net, filename="neat_pole_balancing_winner")
with open('neat_pole_balancing_winner.pkl', 'wb') as output:
pickle.dump(winner_net, output, pickle.HIGHEST_PROTOCOL)
import pickle as pickle
# from __future__ import print_function
import copy
import os
import pandas as pd
from pureples.shared.visualize import draw_net
from sklearn.model_selection import train_test_split
import numpy as np
import visualize
import sklearn.metrics
import neat
# winner_net = neat.nn.RecurrentNetwork.create(winner, config)
filename = 'network_brake'
with open(filename + '.pkl', 'rb') as input:
winner_net = pickle.load(input)
draw_net(winner_net, filename=filename)
neat.reproduction.DefaultReproduction,
neat.species.DefaultSpeciesSet,
neat.stagnation.DefaultStagnation,
'config_cppn_mountain_car')
def run(gens, env):
winner, stats = run_es(gens, env, 200, config, params, sub, max_trials=0)
print("es_hyperneat_mountain_car_small done")
return winner, stats
# If run as script.
if __name__ == '__main__':
# Setup logger and environment.
logger = logging.getLogger()
logger.setLevel(logging.INFO)
env = gym.make("MountainCar-v0")
# Run!
winner = run(200, env)[0]
# Save CPPN if wished reused and draw it + winner to file.
cppn = neat.nn.FeedForwardNetwork.create(winner, config)
network = ESNetwork(sub, cppn, params)
net = network.create_phenotype_network(
filename="es_hyperneat_mountain_car_small_winner")
draw_net(cppn, filename="es_hyperneat_mountain_car_small_cppn")
with open('es_hyperneat_mountain_car_small_cppn.pkl', 'wb') as output:
pickle.dump(cppn, output, pickle.HIGHEST_PROTOCOL)
pop.add_reporter(stats)
pop.add_reporter(neat.reporting.StdOutReporter(True))
winner = pop.run(eval_fitness, gens)
print("es_hyperneat_xor_medium done")
return winner, stats
# If run as script.
if __name__ == '__main__':
winner = run(300)[0]
print('\nBest genome:\n{!s}'.format(winner))
# Verify network output against training data.
print('\nOutput:')
cppn = neat.nn.FeedForwardNetwork.create(winner, config)
network = ESNetwork(sub, cppn, params)
winner_net = network.create_phenotype_network(filename='es_hyperneat_xor_medium_winner.png') # This will also draw winner_net.
for inputs, expected in zip(xor_inputs, xor_outputs):
new_input = inputs + (1.0,)
winner_net.reset()
for i in range(network.activations):
output = winner_net.activate(new_input)
print(" input {!r}, expected output {!r}, got {!r}".format(inputs, expected, output))
# Save CPPN if wished reused and draw it to file.
draw_net(cppn, filename="es_hyperneat_xor_medium_cppn")
with open('es_hyperneat_xor_medium_cppn.pkl', 'wb') as output:
pickle.dump(cppn, output, pickle.HIGHEST_PROTOCOL)
# Create population and train the network. Return winner of network running 100 episodes.
stats_one = neat.statistics.StatisticsReporter()
pop = ini_pop(None, stats_one, config)
pop.run(eval_fitness, gens)
stats_ten = neat.statistics.StatisticsReporter()
pop = ini_pop((pop.population, pop.species, 0), stats_ten, config)
trials = 10
winner_ten = pop.run(eval_fitness, gens)
if max_trials is 0:
return winner_ten, (stats_one, stats_ten)
stats_hundred = neat.statistics.StatisticsReporter()
pop = ini_pop((pop.population, pop.species, 0), stats_hundred, config)
trials = max_trials
winner_hundred = pop.run(eval_fitness, gens)
return winner_hundred, (stats_one, stats_ten, stats_hundred)
logger = logging.getLogger()
logger.setLevel(logging.INFO)
winner = run(100, 1000000, config, params, sub)[0]
cppn = neat.nn.FeedForwardNetwork.create(winner, config)
network = ESNetwork(sub, cppn, params)
net = network.create_phenotype_network(filename="es_hyperneat_2048_winner")
draw_net(cppn, filename="es_hyperNEAT_2048_cppn")
with open('es_hyperNEAT_2048_cppn.pkl', 'wb') as output:
pickle.dump(cppn, output, pickle.HIGHEST_PROTOCOL)
neat.reproduction.DefaultReproduction,
neat.species.DefaultSpeciesSet,
neat.stagnation.DefaultStagnation,
'config_cppn_pole_balancing')
# Use the gym_runner to run this experiment using HyperNEAT.
def run(gens, env):
winner, stats = run_hyper(gens, env, 500, config, sub, activations)
print("hyperneat_polebalancing done")
return winner, stats
# If run as script.
if __name__ == '__main__':
# Setup logger and environment.
logger = logging.getLogger()
logger.setLevel(logging.INFO)
env = gym.make("CartPole-v1")
# Run!
winner = run(100, env)[0]
# Save CPPN if wished reused and draw it + winner to file.
cppn = neat.nn.FeedForwardNetwork.create(winner, config)
winner_net = create_phenotype_network(cppn, sub)
draw_net(cppn, filename="hyperneat_pole_balancing_cppn")
with open('hyperneat_pole_balancing_cppn.pkl', 'wb') as output:
pickle.dump(cppn, output, pickle.HIGHEST_PROTOCOL)
draw_net(winner_net, filename="hyperneat_pole_balancing_winner")
def run(gens, env):
"""
Run the pole balancing task using the Gym environment
Returns the winning genome and the statistics of the run.
"""
winner, stats = run_hyper(gens, env, 500, CONFIG, SUBSTRATE, ACTIVATIONS)
print("hyperneat_polebalancing done")
return winner, stats
# If run as script.
if __name__ == '__main__':
# Setup logger and environment.
LOGGER = logging.getLogger()
LOGGER.setLevel(logging.INFO)
ENVIRONMENT = gym.make("CartPole-v1")
# Run! Only relevant to look at the winner.
WINNER = run(100, ENVIRONMENT)[0]
# Save CPPN if wished reused and draw it + winner to file.
CPPN = neat.nn.FeedForwardNetwork.create(WINNER, CONFIG)
WINNER_NET = create_phenotype_network(CPPN, SUBSTRATE)
draw_net(
CPPN, filename="pureples/experiments/pole_balancing/hyperneat_pole_balancing_cppn")
with open('pureples/experiments/pole_balancing/hyperneat_pole_balancing_cppn.pkl', 'wb') as output:
pickle.dump(CPPN, output, pickle.HIGHEST_PROTOCOL)
draw_net(
WINNER_NET, filename="pureples/experiments/pole_balancing/hyperneat_pole_balancing_winner")
#winner = pop.run(eval_fitness, gens)
try:
winner = pop.run(eval_parallel, gens)
return winner, stats
except multiprocessing.context.TimeoutError:
os.system('make die')
os._exit()
def run(gens):
winner, stats = run_hyper(gens, config, sub, activations)
print("hyperneat_mario done")
return winner, stats
if __name__ == '__main__':
# Setup logger and environment.
logger = logging.getLogger()
logger.setLevel(logging.INFO)
# Run!
winner = run(700)[0]
# Save CPPN and draw it + winner to file.
cppn = neat.nn.FeedForwardNetwork.create(winner, config)
net = create_phenotype_network(cppn, sub)
draw_net(cppn, filename="hyperneat_mario_cppn")
draw_net(net, filename="hyperneat_mario_winner")
with open('hyperneat_mario_cppn.pkl', 'wb') as output:
pickle.dump(cppn, output, pickle.HIGHEST_PROTOCOL)
winner, stats = run_es(gens,
env,
200,
CONFIG,
params(version),
SUBSTRATE,
max_trials=0)
print(f"es_hyperneat_mountain_car_{VERSION_TEXT} done")
return winner, stats
# If run as script.
if __name__ == '__main__':
# Setup logger and environment.
LOGGER = logging.getLogger()
LOGGER.setLevel(logging.INFO)
ENVIRONMENT = gym.make("MountainCar-v0")
# Run! Only relevant to look at the winner.
WINNER = run(200, ENVIRONMENT, VERSION)[0]
# Save CPPN if wished reused and draw it + winner to file.
CPPN = neat.nn.FeedForwardNetwork.create(WINNER, CONFIG)
NETWORK = ESNetwork(SUBSTRATE, CPPN, params)
NET = NETWORK.create_phenotype_network(
filename=f"es_hyperneat_mountain_car_{VERSION_TEXT}_winner")
draw_net(CPPN, filename=f"es_hyperneat_mountain_car_{VERSION_TEXT}_cppn")
with open(f'es_hyperneat_mountain_car_{VERSION_TEXT}_cppn.pkl',
'wb') as output:
pickle.dump(CPPN, output, pickle.HIGHEST_PROTOCOL)
winner = pop.run(eval_fitness, gens)
print("hyperneat_xor done")
return winner, stats
# If run as script.
if __name__ == '__main__':
WINNER = run(300)[0] # Only relevant to look at the winner.
print('\nBest genome:\n{!s}'.format(WINNER))
# Verify network output against training data.
print('\nOutput:')
CPPN = neat.nn.FeedForwardNetwork.create(WINNER, CONFIG)
WINNER_NET = create_phenotype_network(CPPN, SUBSTRATE)
for inputs, expected in zip(XOR_INPUTS, XOR_OUTPUTS):
new_input = inputs + (1.0,)
WINNER_NET.reset()
for i in range(ACTIVATIONS):
output = WINNER_NET.activate(new_input)
print(" input {!r}, expected output {!r}, got {!r}".format(
inputs, expected, output))
# Save CPPN if wished reused and draw it to file along with the winner.
with open('pureples/experiments/xor/hyperneat_xor_cppn.pkl', 'wb') as output:
pickle.dump(CPPN, output, pickle.HIGHEST_PROTOCOL)
draw_net(CPPN, filename="pureples/experiments/xor/hyperneat_xor_cppn")
draw_net(WINNER_NET, filename="pureples/experiments/xor/hyperneat_xor_winner")
# fitness function.
def run(gens):
pop = neat.population.Population(config)
stats = neat.statistics.StatisticsReporter()
pop.add_reporter(stats)
pop.add_reporter(neat.reporting.StdOutReporter(True))
winner = pop.run(eval_fitness, gens)
print("neat_xor done")
return winner, stats
# If run as script.
if __name__ == '__main__':
winner = run(300)[0]
print('\nBest genome:\n{!s}'.format(winner))
# Verify network output against training data.
print('\nOutput:')
winner_net = neat.nn.FeedForwardNetwork.create(winner, config)
for inputs, expected in zip(xor_inputs, xor_outputs):
new_input = inputs + (1.0, )
output = winner_net.activate(new_input)
print(" input {!r}, expected output {!r}, got {!r}".format(
inputs, expected, output))
# Save net if wished reused and draw it to a file.
with open('winner_neat_xor.pkl', 'wb') as output:
pickle.dump(winner_net, output)
draw_net(winner_net, filename="neat_xor_winner")
pop.add_reporter(stats)
pop.add_reporter(neat.reporting.StdOutReporter(True))
winner = pop.run(eval_fitness, gens)
print("es_hyperneat_xor_large done")
return winner, stats
# If run as script.
if __name__ == '__main__':
winner = run(300)[0]
print('\nBest genome:\n{!s}'.format(winner))
# Verify network output against training data.
print('\nOutput:')
cppn = neat.nn.FeedForwardNetwork.create(winner, config)
network = ESNetwork(sub, cppn, params)
winner_net = network.create_phenotype_network(filename='es_hyperneat_xor_large_winner.png') # This will also draw winner_net.
for inputs, expected in zip(xor_inputs, xor_outputs):
new_input = inputs + (1.0,)
winner_net.reset()
for i in range(network.activations):
output = winner_net.activate(new_input)
print(" input {!r}, expected output {!r}, got {!r}".format(inputs, expected, output))
# Save CPPN if wished reused and draw it to file.
draw_net(cppn, filename="es_hyperneat_xor_large_cppn")
with open('es_hyperneat_xor_large_cppn.pkl', 'wb') as output:
pickle.dump(cppn, output, pickle.HIGHEST_PROTOCOL)
print(f"es_hyperneat_polebalancing_{VERSION_TEXT} done")
return winner, stats
# If run as script.
if __name__ == '__main__':
# Setup logger and environment.
LOGGER = logging.getLogger()
LOGGER.setLevel(logging.INFO)
ENVIRONMENT = gym.make("CartPole-v1")
# Run! Only relevant to look at the winner.
WINNER = run(100, ENVIRONMENT, VERSION)[0]
# Save CPPN if wished reused and draw it + winner to file.
CPPN = neat.nn.FeedForwardNetwork.create(WINNER, CONFIG)
NETWORK = ESNetwork(SUBSTRATE, CPPN, params(VERSION))
NET = NETWORK.create_phenotype_network(
filename=
f"pureples/experiments/pole_balancing/es_hyperneat_pole_balancing_{VERSION_TEXT}_winner"
)
draw_net(
CPPN,
filename=
f"pureples/experiments/pole_balancing/es_hyperneat_pole_balancing_{VERSION_TEXT}_cppn"
)
with open(
f'pureples/experiments/pole_balancing/es_hyperneat_pole_balancing_{VERSION_TEXT}_cppn.pkl',
'wb') as output:
pickle.dump(CPPN, output, pickle.HIGHEST_PROTOCOL)
neat.species.DefaultSpeciesSet,
neat.stagnation.DefaultStagnation,
'config_cppn_pole_balancing')
# Use the gym_runner to run this experiment using ES-HyperNEAT.
def run(gens, env):
winner, stats = run_es(gens, env, 500, config, params, sub)
print("es_hyperneat_polebalancing_large done")
return winner, stats
# If run as script.
if __name__ == '__main__':
# Setup logger and environment.
logger = logging.getLogger()
logger.setLevel(logging.INFO)
env = gym.make("CartPole-v1")
# Run!
winner = run(100, env)[0]
# Save CPPN if wished reused and draw it + winner to file.
cppn = neat.nn.FeedForwardNetwork.create(winner, config)
network = ESNetwork(sub, cppn, params)
net = network.create_phenotype_network(
filename="es_hyperneat_pole_balancing_large_winner")
draw_net(cppn, filename="es_hyperneat_pole_balancing_large_cppn")
with open('es_hyperneat_pole_balancing_large_cppn.pkl', 'wb') as output:
pickle.dump(cppn, output, pickle.HIGHEST_PROTOCOL)
pop.add_reporter(stats)
pop.add_reporter(neat.reporting.StdOutReporter(True))
winner = pop.run(eval_fitness, gens)
print("es_hyperneat_xor_small done")
return winner, stats
# If run as script.
if __name__ == '__main__':
winner = run(300)[0]
print('\nBest genome:\n{!s}'.format(winner))
# Verify network output against training data.
print('\nOutput:')
cppn = neat.nn.FeedForwardNetwork.create(winner, config)
network = ESNetwork(sub, cppn, params)
winner_net = network.create_phenotype_network(filename='es_hyperneat_xor_small_winner.png') # This will also draw winner_net.
for inputs, expected in zip(xor_inputs, xor_outputs):
new_input = inputs + (1.0,)
winner_net.reset()
for i in range(network.activations):
output = winner_net.activate(new_input)
print(" input {!r}, expected output {!r}, got {!r}".format(inputs, expected, output))
# Save CPPN if wished reused and draw it to file.
draw_net(cppn, filename="es_hyperneat_xor_small_cppn")
with open('es_hyperneat_xor_small_cppn.pkl', 'wb') as output:
pickle.dump(cppn, output, pickle.HIGHEST_PROTOCOL)
# Config for NEAT.
config = neat.config.Config(neat.genome.DefaultGenome,
neat.reproduction.DefaultReproduction,
neat.species.DefaultSpeciesSet,
neat.stagnation.DefaultStagnation,
'config_neat_mountain_car')
def run(gens, env):
winner, stats = run_neat(gens, env, 200, config, max_trials=0)
print("neat_mountain_car done")
return winner, stats
# If run as script.
if __name__ == '__main__':
# Setup logger and environment.
logger = logging.getLogger()
logger.setLevel(logging.INFO)
env = gym.make("MountainCar-v0")
# Run!
winner = run(20, env)[0]
# Save net if wished reused and draw it to file.
net = neat.nn.RecurrentNetwork.create(winner, config)
draw_net(net, filename="neat_mountain_car_winner")
with open('neat_mountain_car_winner.pkl', 'wb') as output:
pickle.dump(net, output, pickle.HIGHEST_PROTOCOL)
pop.add_reporter(neat.reporting.StdOutReporter(True))
winner = pop.run(task.eval_fitness, gens)
print("es trade god summoned")
return winner, stats
# If run as script.
if __name__ == '__main__':
task = PurpleTrader(8)
#print(task.trial_run())
winner = run_pop(task, 89)[0]
print('\nBest genome:\n{!s}'.format(winner))
# Verify network output against training data.
print('\nOutput:')
[cppn] = create_cppn(winner, task.config, task.leaf_names, ['cppn_out'])
network = ESNetwork(task.subStrate, cppn, task.params)
with open('es_trade_god_cppn_3d.pkl', 'wb') as output:
pickle.dump(winner, output)
#draw_net(cppn, filename="es_trade_god")
winner_net = network.create_phenotype_network_nd(
'dabestest.png') # This will also draw winner_net.
# Save CPPN if wished reused and draw it to file.
draw_net(cppn, filename="es_trade_god")
'''
for x in range(len(task.hs.hist_shaped[0])):
print(task.hs.hist_shaped[1][x][3],task.hs.hist_shaped[0][x][3])
'''
pop.add_reporter(stats)
pop.add_reporter(neat.reporting.StdOutReporter(True))
winner = pop.run(eval_fitness, gens)
print("hyperneat_xor done")
return winner, stats
# If run as script.
if __name__ == '__main__':
winner = run(300)[0]
print('\nBest genome:\n{!s}'.format(winner))
# Verify network output against training data.
print('\nOutput:')
cppn = neat.nn.FeedForwardNetwork.create(winner, config)
winner_net = create_phenotype_network(cppn, sub)
for inputs, expected in zip(xor_inputs, xor_outputs):
new_input = inputs + (1.0, )
winner_net.reset()
for i in range(activations):
output = winner_net.activate(new_input)
print(" input {!r}, expected output {!r}, got {!r}".format(
inputs, expected, output))
# Save CPPN if wished reused and draw it to file along with the winner.
with open('hyperneat_xor_cppn.pkl', 'wb') as output:
pickle.dump(cppn, output, pickle.HIGHEST_PROTOCOL)
draw_net(cppn, filename="hyperneat_xor_cppn")
draw_net(winner_net, filename="hyperneat_xor_winner")
neat.reproduction.DefaultReproduction,
neat.species.DefaultSpeciesSet,
neat.stagnation.DefaultStagnation,
'config_cppn_mountain_car')
def run(gens, env):
winner, stats = run_es(gens, env, 200, config, params, sub, max_trials=0)
print("es_hyperneat_mountain_car_large done")
return winner, stats
# If run as script.
if __name__ == '__main__':
# Setup logger and environment.
logger = logging.getLogger()
logger.setLevel(logging.INFO)
env = gym.make("MountainCar-v0")
# Run!
winner = run(200, env)[0]
# Save CPPN if wished reused and draw it + winner to file.
cppn = neat.nn.FeedForwardNetwork.create(winner, config)
network = ESNetwork(sub, cppn, params)
net = network.create_phenotype_network(
filename="es_hyperneat_mountain_car_large_winner")
draw_net(cppn, filename="es_hyperneat_mountain_car_large_cppn")
with open('es_hyperneat_mountain_car_large_cppn.pkl', 'wb') as output:
pickle.dump(cppn, output, pickle.HIGHEST_PROTOCOL)
neat.species.DefaultSpeciesSet,
neat.stagnation.DefaultStagnation,
'config_cppn_pole_balancing')
# Use the gym_runner to run this experiment using ES-HyperNEAT.
def run(gens, env):
winner, stats = run_es(gens, env, 500, config, params, sub)
print("es_hyperneat_polebalancing_medium done")
return winner, stats
# If run as script.
if __name__ == '__main__':
# Setup logger and environment.
logger = logging.getLogger()
logger.setLevel(logging.INFO)
env = gym.make("CartPole-v1")
# Run!
winner = run(100, env)[0]
# Save CPPN if wished reused and draw it + winner to file.
cppn = neat.nn.FeedForwardNetwork.create(winner, config)
network = ESNetwork(sub, cppn, params)
net = network.create_phenotype_network(
filename="es_hyperneat_pole_balancing_medium_winner")
draw_net(cppn, filename="es_hyperneat_pole_balancing_medium_cppn")
with open('es_hyperneat_pole_balancing_medium_cppn.pkl', 'wb') as output:
pickle.dump(cppn, output)
neat.reproduction.DefaultReproduction,
neat.species.DefaultSpeciesSet,
neat.stagnation.DefaultStagnation,
'config_cppn_mountain_car')
def run(gens, env):
winner, stats = run_es(gens, env, 200, config, params, sub, max_trials=0)
print("es_hyperneat_mountain_car_medium done")
return winner, stats
# If run as script.
if __name__ == '__main__':
# Setup logger and environment.
logger = logging.getLogger()
logger.setLevel(logging.INFO)
env = gym.make("MountainCar-v0")
# Run!
winner = run(200, env)[0]
# Save CPPN if wished reused and draw it + winner to file.
cppn = neat.nn.FeedForwardNetwork.create(winner, config)
network = ESNetwork(sub, cppn, params)
net = network.create_phenotype_network(
filename="es_hyperneat_mountain_car_medium_winner")
draw_net(cppn, filename="es_hyperneat_mountain_car_medium_cppn")
with open('es_hyperneat_mountain_car_medium_cppn.pkl', 'wb') as output:
pickle.dump(cppn, output, pickle.HIGHEST_PROTOCOL)
Returns the winning genome and the statistics of the run.
"""
winner, stats = run_hyper(gens,
env,
200,
CONFIG,
SUBSTRATE,
ACTIVATIONS,
max_trials=0)
print("hyperneat_mountain_car done")
return winner, stats
# If run as script.
if __name__ == '__main__':
# Setup logger and environment.
LOGGER = logging.getLogger()
LOGGER.setLevel(logging.INFO)
ENVIRONMENT = gym.make("MountainCar-v0")
# Run! Only relevant to look at the winner.
WINNER = run(200, ENVIRONMENT)[0]
# Save CPPN if wished reused and draw it + winner to file.
cppn = neat.nn.FeedForwardNetwork.create(WINNER, CONFIG)
NET = create_phenotype_network(cppn, SUBSTRATE)
draw_net(cppn, filename="hyperneat_mountain_car_cppn")
draw_net(NET, filename="hyperneat_mountain_car_winner")
with open('hyperneat_mountain_car_cppn.pkl', 'wb') as output:
pickle.dump(cppn, output, pickle.HIGHEST_PROTOCOL)
config = neat.config.Config(neat.genome.DefaultGenome, neat.reproduction.DefaultReproduction,
neat.species.DefaultSpeciesSet, neat.stagnation.DefaultStagnation,
'config_cppn_pole_balancing')
# Use the gym_runner to run this experiment using ES-HyperNEAT.
def run(gens, env):
winner, stats = run_es(gens, env, 500, config, params, sub)
print("es_hyperneat_polebalancing_small done")
return winner, stats
# If run as script.
if __name__ == '__main__':
# Setup logger and environment.
logger = logging.getLogger()
logger.setLevel(logging.INFO)
env = gym.make("CartPole-v1")
# Run!
winner = run(100, env)[0]
# Save CPPN if wished reused and draw it + winner to file.
cppn = neat.nn.FeedForwardNetwork.create(winner, config)
network = ESNetwork(sub, cppn, params)
net = network.create_phenotype_network(filename="es_hyperneat_pole_balancing_small_winner")
draw_net(cppn, filename="es_hyperneat_pole_balancing_small_cppn")
with open('es_hyperneat_pole_balancing_small_cppn.pkl', 'wb') as output:
pickle.dump(cppn, output, pickle.HIGHEST_PROTOCOL)
WINNER = run(300, VERSION)[0] # Only relevant to look at the winner.
print('\nBest genome:\n{!s}'.format(WINNER))
# Verify network output against training data.
print('\nOutput:')
CPPN = neat.nn.FeedForwardNetwork.create(WINNER, CONFIG)
NETWORK = ESNetwork(SUBSTRATE, CPPN, DYNAMIC_PARAMS)
# This will also draw winner_net.
WINNER_NET = NETWORK.create_phenotype_network(
filename=
f'pureples/experiments/xor/es_hyperneat_xor_{VERSION_TEXT}_winner.png')
for inputs, expected in zip(XOR_INPUTS, XOR_OUTPUTS):
new_input = inputs + (1.0, )
WINNER_NET.reset()
for i in range(NETWORK.activations):
output = WINNER_NET.activate(new_input)
print(" input {!r}, expected output {!r}, got {!r}".format(
inputs, expected, output))
# Save CPPN if wished reused and draw it to file.
draw_net(CPPN,
filename=
f"pureples/experiments/xor/es_hyperneat_xor_{VERSION_TEXT}_cppn")
with open(
f'pureples/experiments/xor/es_hyperneat_xor_{VERSION_TEXT}_cppn.pkl',
'wb') as output:
pickle.dump(CPPN, output, pickle.HIGHEST_PROTOCOL)
"""
pop = neat.population.Population(CONFIG)
stats = neat.statistics.StatisticsReporter()
pop.add_reporter(stats)
pop.add_reporter(neat.reporting.StdOutReporter(True))
winner = pop.run(eval_fitness, gens)
print("neat_xor done")
return winner, stats
# If run as script.
if __name__ == '__main__':
WINNER = run(300)[0] # Only relevant to look at the winner.
print('\nBest genome:\n{!s}'.format(WINNER))
# Verify network output against training data.
print('\nOutput:')
WINNER_NET = neat.nn.FeedForwardNetwork.create(WINNER, CONFIG)
for inputs, expected in zip(XOR_INPUTS, XOR_OUTPUTS):
new_input = inputs + (1.0, )
output = WINNER_NET.activate(new_input)
print(" input {!r}, expected output {!r}, got {!r}".format(
inputs, expected, output))
# Save net if wished reused and draw it to a file.
with open('pureples/experiments/xor/winner_neat_xor.pkl', 'wb') as output:
pickle.dump(WINNER_NET, output, pickle.HIGHEST_PROTOCOL)
draw_net(WINNER_NET, filename="pureples/experiments/xor/neat_xor_winner")
