class Experiment:
def __init__(self, args):
self._args = args
def generate_map(self):
if self._args.model:
self.map_points = self._load_map_points_from_model()
else:
self.map_points = self._generate_random_map_points()
def _load_map_points_from_model(self):
model = storage.load(self._args.model)[0]
return model.normalized_observed_reductions
def _generate_random_map_points(self):
samples, _labels = make_classification(
n_features=2, n_redundant=0, n_informative=1,
n_clusters_per_class=1, n_samples=1000)
return self._normalize(samples)
def _normalize(self, points):
min_value = min([min(point) for point in points])
max_value = max([max(point) for point in points])
return (points - min_value) / (max_value - min_value)
def create_navigator(self):
self.navigator = Navigator(map_points=self.map_points)
def generate_new_path(self):
departure = random.choice(self.map_points)
self._generate_path(departure)
def extend_path(self):
departure = self.path[-1]
self._generate_path(departure)
def _get_slider_value(self, name):
return float(self.window.sliders[name].value()) / SLIDER_PRECISION
def _generate_path(self, departure):
self.path_segments = self.navigator.generate_path(
departure,
num_segments=10,
novelty=self._get_slider_value("novelty"),
extension=self._get_slider_value("extension"),
location_preference=self._get_slider_value("location_preference"))
self.path = interpolation.interpolate(
self.path_segments,
resolution=100)
self.path_followers = [
PathFollower(self.path, dynamics=dynamics.constant_dynamics()),
PathFollower(self.path, dynamics=dynamics.sine_dynamics()),
]
def proceed(self, time_increment):
for path_follower in self.path_followers:
path_follower.proceed(time_increment * VELOCITY)
