def test_non_gradient_can_set_bounds_range_builder_properties(
model: Model) -> None:
non_gradient_optimizer = NonGradientOptimizer(model.a, bounds_range=0.1)
logProb = non_gradient_optimizer.max_a_posteriori()
sum_ab = model.a.get_value() + model.b.get_value()
assert not (19.9 < sum_ab < 20.1)
def test_max_likelihood_non_gradient(model: Model) -> None:
non_gradient_optimizer = NonGradientOptimizer(model.a)
logProb = non_gradient_optimizer.max_likelihood()
assert logProb < 0.
sum_ab = model.a.get_value() + model.b.get_value()
assert 19.9 < sum_ab < 20.1
def test_max_likelihood_non_gradient(model: Model) -> None:
non_gradient_optimizer = NonGradientOptimizer(model.a)
result = non_gradient_optimizer.max_likelihood()
assert result.fitness() < 0.
sum_ab = result.value_for(model.a) + result.value_for(model.b)
assert 19.9 < sum_ab < 20.1
def test_non_gradient_can_set_max_eval_builder_properties(
model: Model) -> None:
non_gradient_optimizer = NonGradientOptimizer(model.a, max_evaluations=5)
with pytest.raises(Py4JJavaError):
#This throws a Gradient Optimizer: "Reached Max Evaluations" error
logProb = non_gradient_optimizer.max_a_posteriori()
def test_non_gradient_can_set_max_eval_builder_properties(
model: Model) -> None:
non_gradient_optimizer = NonGradientOptimizer(model.a,
BOBYQA(max_evaluations=5))
with pytest.raises(
Py4JJavaError,
match=r"An error occurred while calling o[\d]*.maxAPosteriori."):
# This throws a Gradient Optimizer: "Reached Max Evaluations" error
logProb = non_gradient_optimizer.max_a_posteriori()
def non_gradient_optimizer_example():
model = build_model()
model.first_thermometer.observe(25)
bayes_net = model.to_bayes_net()
# %%SNIPPET_START%% PythonNonGradientOptimizer
optimizer = NonGradientOptimizer(bayes_net,
max_evaluations=5000,
bounds_range=100000.,
initial_trust_region_radius=5.,
stopping_trust_region_radius=2e-8)
optimizer.max_a_posteriori()
calculated_temperature = model.temperature.get_value()
def test_non_gradient_op_throws_with_invalid_net_param() -> None:
with pytest.raises(
TypeError,
match=r"net must be a Vertex or a BayesNet. Was given {}".format(
int)):
NonGradientOptimizer(
500) # type: ignore # this is expected to fail mypy
def test_non_gradient_op_vertex(model: Model) -> None:
non_gradient_optimizer = NonGradientOptimizer(model.a)
assert len(list(non_gradient_optimizer.net.get_latent_vertices())) == 2
def test_non_gradient_op_bayes_net(model: Model) -> None:
net = BayesNet(model.a.get_connected_graph())
gradient_optimizer = NonGradientOptimizer(net)
assert gradient_optimizer.net is net