def __init__(
self,
model: Callable = lambda x: NormalContinuous(bounds=x.bounds),
accumulate_rewards: bool = False,
minimize: bool = True,
start_same_pos: bool = False,
*args,
**kwargs,
):
"""
Initialize a :class:`FunctionMapper`.
Args:
model: A function that returns an instance of a :class:`Model`.
accumulate_rewards: If ``True`` the rewards obtained after transitioning \
to a new state will accumulate. If ``False`` only the last \
reward will be taken into account.
minimize: If ``True`` the algorithm will perform a minimization \
process. If ``False`` it will be a maximization process.
start_same_pos: If ``True`` all the walkers will have the same \
starting position.
*args: Passed :class:`Swarm` __init__.
**kwargs: Passed :class:`Swarm` __init__.
"""
super(FunctionMapper,
self).__init__(model=model,
accumulate_rewards=accumulate_rewards,
minimize=minimize,
*args,
**kwargs)
self.start_same_pos = start_same_pos
def create_model(name="discrete"):
if name == "discrete":
return lambda: DiscreteUniform(n_actions=10)
elif name == "continuous":
bs = Bounds(low=-1, high=1, shape=(3, ))
return lambda: ContinuousUniform(bounds=bs)
elif name == "random_normal":
bs = Bounds(low=-1, high=1, shape=(3, ))
return lambda: NormalContinuous(loc=0, scale=1, bounds=bs)
raise ValueError("Invalid param `name`.")
def create_function_swarm():
env = Rastrigin(dims=2)
swarm = FunctionMapper(
model=lambda x: NormalContinuous(bounds=env.bounds),
env=lambda: env,
n_walkers=5,
max_epochs=5,
prune_tree=True,
reward_scale=2,
minimize=False,
)
return swarm
def test_sample(self):
bounds = Bounds(low=-5, high=5, shape=(3, ))
model = NormalContinuous(bounds=bounds)
actions = model.predict(batch_size=10000).actions
assert actions.min() >= -5
assert actions.max() <= 5
assert numpy.allclose(actions.mean(), 0, atol=0.05)
assert numpy.allclose(actions.std(), 1, atol=0.05)
bounds = Bounds(low=-10, high=30, shape=(3, 10))
model = NormalContinuous(bounds=bounds, loc=5, scale=2)
actions = model.predict(batch_size=10000).actions
assert actions.min() >= -10
assert actions.max() <= 30
assert numpy.allclose(actions.mean(), 5, atol=0.05), actions.mean()
assert numpy.allclose(actions.std(), 2, atol=0.05), actions.std()
def gaussian_model(env):
# Gaussian of mean 0 and std of 10, adapted to the environment bounds
return NormalContinuous(scale=10, loc=0.0, bounds=env.bounds)