def _calculate_solution_value(self, solution, data_set, learner):
X = get_input_variables(data_set).values
target = get_target_variable(data_set).values
neural_network = FeedForwardNetwork.create(solution,
learner.configuration)
prediction = self._predict_neural_network(neural_network, X)
return self.metric.evaluate(prediction, target)
def _predict_genome(self, genome):
neural_network = FeedForwardNetwork.create(genome, self.configuration)
predictions = array([])
for data in self.input_matrix:
predictions = append(predictions,
float(neural_network.activate(data)[0]))
return predictions
def predict(self, input_matrix):
neural_network = FeedForwardNetwork.create(self.champion,
self.configuration)
predictions = array([])
for data in input_matrix:
predictions = append(predictions,
float(neural_network.activate(data)[0]))
return predictions
def run(n_generations):
# Load the config file, which is assumed to live in
# the same directory as this script.
config_path = os.path.join(os.path.dirname(__file__), "neat.cfg")
config = neat.Config(
neat.DefaultGenome,
neat.DefaultReproduction,
neat.DefaultSpeciesSet,
neat.DefaultStagnation,
config_path,
)
evaluator = MultiEnvEvaluator(make_net,
activate_net,
make_env=make_env,
max_env_steps=max_env_steps)
def eval_genomes(genomes, config):
for _, genome in genomes:
genome.fitness = evaluator.eval_genome(genome, config)
pop = neat.Population(config)
stats = neat.StatisticsReporter()
pop.add_reporter(stats)
reporter = neat.StdOutReporter(True)
pop.add_reporter(reporter)
logger = LogReporter("neat_local.log", evaluator.eval_genome)
pop.add_reporter(logger)
winner = pop.run(eval_genomes, n_generations)
node_names = {
-1: 'Cart Position',
-2: 'Cart Velocity',
-3: 'Pole Angle',
-4: 'Pole Velocity At Tip',
0: 'Push cart to the left or to the right'
}
# Visualization
visualize.draw_net(config,
winner,
True,
filename="graph_neat_examples_CartPole-v1",
node_names=node_names)
visualize.plot_stats(stats,
ylog=False,
view=True,
filename="stats_neat_examples_CartPole-v1")
visualize.plot_species(stats,
view=True,
filename="species_neat_examples_CartPole-v1")
winner_net = FeedForwardNetwork.create(winner, config)
return winner_net
def develop_phenotype(genome, config):
pooling_net = FeedForwardNetwork.create(
genome=genome.pooling_genome,
config=POOL_CONFIG,
)
decision_net = FeedForwardNetwork.create(
genome=genome.decision_genome,
config=DECISION_CONFIG,
)
memoized_pooling_activation = lru_cache()(pooling_net.activate)
def brain(inputs):
pooled_layer = [
memoized_pooling_activation(seq)[0] for seq in inputs
]
decision_layer = decision_net.activate(pooled_layer)
return decision_layer
return brain
def reset(self, genome, config):
self.net = FeedForwardNetwork.create(genome, config)