def mapper():
def potential_well(x):
return numpy.sum((x - 1)**2, 1) - 1
bounds = Bounds.from_tuples([(-10, 10), (-5, 5)])
def model(x):
return NormalContinuous(bounds=bounds, env=x)
return FunctionMapper.from_function(
n_vectors=5,
function=potential_well,
bounds=bounds,
model=model,
shape=(2, ),
n_walkers=10,
reward_scale=1,
accumulate_rewards=False,
)
def create_model(name="es_model"):
if name == "es_model":
bs = Bounds(low=-10, high=10, shape=(BATCH_SIZE,))
return lambda: ESModel(bounds=bs)
raise ValueError("Invalid param `name`.")
def custom_domain_function():
bounds = Bounds(shape=(2,), high=10, low=-5, dtype=float)
env = Function(
function=sphere, bounds=bounds, custom_domain_check=lambda x: numpy.linalg.norm(x) < 5.0
)
return env
def local_minimizer():
bounds = Bounds(shape=(2,), high=10, low=-5, dtype=float)
env = Function(function=sphere, bounds=bounds)
return MinimizerWrapper(env)
def test_minimizer_getattr(self):
bounds = Bounds(shape=(2,), high=10, low=-5, dtype=float)
env = Function(function=sphere, bounds=bounds)
minim = MinimizerWrapper(env)
assert minim.shape == env.shape
def det_cmaes():
bs = Bounds(low=-10, high=10, shape=(5, ))
return DeterministicCMAES(bounds=bs, sigma=0.3)
def cmaes():
bs = Bounds(low=-10, high=10, shape=(5, ))
return CMAES(bounds=bs, sigma=0.3)