def test_can_pass_scalar_to_vertex(jvm_view: JVMView) -> None:
gaussian = Vertex(jvm_view.GaussianVertex, 0., 1.)
sample = gaussian.sample()
assert type(sample) == numpy_types
assert sample.shape == ()
assert sample.dtype == float
def test_get_vertex_id(jvm_view: JVMView) -> None:
gaussian = Vertex(jvm_view.GaussianVertex, 0., 1.)
java_id = gaussian.unwrap().getId().getValue()
python_id = gaussian.get_id()
assert all(value in python_id for value in java_id)
def test_can_pass_vertex_to_vertex(jvm_view: JVMView) -> None:
mu = Gaussian(0., 1.)
gaussian = Vertex(jvm_view.GaussianVertex, "gaussian", mu, Const(1.))
sample = gaussian.sample()
assert type(sample) == numpy_types
assert sample.shape == ()
assert sample.dtype == float
def test_java_collections_to_generator() -> None:
gaussian = Gaussian(0., 1.)
java_collections = gaussian.unwrap().getConnectedGraph()
python_list = list(Vertex._to_generator(java_collections))
java_vertex_ids = [
Vertex._get_python_id(java_vertex) for java_vertex in java_collections
]
assert java_collections.size() == len(python_list)
assert all(
type(element) == Double and element.get_id() in java_vertex_ids
for element in python_list)
def _samples_generator(sample_iterator: JavaObject,
vertices_unwrapped: JavaList, live_plot: bool,
refresh_every: int, ax: Any) -> sample_generator_types:
traces = []
x0 = 0
while (True):
network_sample = sample_iterator.next()
sample = {
Vertex._get_python_label(vertex_unwrapped):
Tensor._to_ndarray(network_sample.get(vertex_unwrapped))
for vertex_unwrapped in vertices_unwrapped
}
if live_plot:
traces.append(sample)
if len(traces) % refresh_every == 0:
joined_trace = {
k: [t[k] for t in traces]
for k in sample.keys()
}
if ax is None:
ax = traceplot(joined_trace, x0=x0)
else:
traceplot(joined_trace, ax=ax, x0=x0)
x0 += refresh_every
traces = []
yield sample
def assert_vertex_value_equals_scalar(vertex: Vertex, expected_type: Type,
scalar: primitive_types) -> None:
vertex_value = vertex.get_value()
assert vertex_value == scalar
assert type(vertex_value) == numpy_types
assert vertex_value.shape == ()
assert vertex_value.dtype == expected_type
def assert_vertex_value_equals_pandas(vertex: Vertex, expected_type: Type,
pandas: pandas_types) -> None:
get_value = vertex.get_value()
expected_value = pandas.values.astype(expected_type).reshape(
get_value.shape)
assert np.array_equal(get_value, expected_value)
assert np.issubdtype(get_value.dtype, expected_type)
def sample(net: BayesNet,
sample_from: Iterable[Vertex],
algo: str = 'metropolis',
proposal_distribution: str = None,
proposal_distribution_sigma: numpy_types = None,
proposal_listeners=[],
draws: int = 500,
drop: int = 0,
down_sample_interval: int = 1,
plot: bool = False,
ax: Any = None) -> sample_types:
sampling_algorithm: JavaObject = build_sampling_algorithm(
algo, proposal_distribution, proposal_distribution_sigma,
proposal_listeners)
vertices_unwrapped: JavaList = k.to_java_object_list(sample_from)
network_samples: JavaObject = sampling_algorithm.getPosteriorSamples(
net.unwrap(), vertices_unwrapped,
draws).drop(drop).downSample(down_sample_interval)
vertex_samples = {
Vertex._get_python_label(vertex_unwrapped): list(
map(Tensor._to_ndarray,
network_samples.get(vertex_unwrapped).asList()))
for vertex_unwrapped in vertices_unwrapped
}
if plot:
traceplot(vertex_samples, ax=ax)
return vertex_samples
def _samples_generator(
sample_iterator: JavaObject, vertices_unwrapped: JavaList,
live_plot: bool, refresh_every: int, ax: Any, all_scalar: bool,
id_to_label: Dict[Tuple[int, ...], str]) -> sample_generator_types:
traces = []
x0 = 0
while (True):
network_sample = sample_iterator.next()
if all_scalar:
sample: sample_generator_dict_type = {
id_to_label[Vertex._get_python_id(vertex_unwrapped)]:
Tensor._to_ndarray(
network_sample.get(vertex_unwrapped)).item()
for vertex_unwrapped in vertices_unwrapped
}
else:
sample = __create_multi_indexed_samples_generated(
vertices_unwrapped, network_sample, id_to_label)
if live_plot:
traces.append(sample)
if len(traces) % refresh_every == 0:
joined_trace: sample_types = {
k: [t[k] for t in traces]
for k in sample.keys()
}
if ax is None:
ax = traceplot(joined_trace, x0=x0)
else:
traceplot(joined_trace, ax=ax, x0=x0)
x0 += refresh_every
traces = []
yield sample
def __create_single_indexed_samples(
network_samples: JavaObject, vertices_unwrapped: JavaList,
id_to_label: Dict[Tuple[int, ...], str]) -> sample_types:
vertex_samples: sample_types = {}
for vertex_unwrapped in vertices_unwrapped:
vertex_label = id_to_label[Vertex._get_python_id(vertex_unwrapped)]
samples_for_vertex = __get_vertex_samples(network_samples,
vertex_unwrapped)
vertex_samples[vertex_label] = samples_for_vertex.tolist()
return vertex_samples
def _samples_generator(sample_iterator: JavaObject,
vertices_unwrapped: JavaList) -> sample_generator_types:
while (True):
network_sample = sample_iterator.next()
sample = {
Vertex._get_python_id(vertex_unwrapped):
Tensor._to_ndarray(network_sample.get(vertex_unwrapped))
for vertex_unwrapped in vertices_unwrapped
}
yield sample
def test_cannot_pass_generic_to_vertex(jvm_view: JVMView) -> None:
class GenericExampleClass:
pass
with pytest.raises(
ValueError,
match=r"Can't parse generic argument. Was given {}".format(
GenericExampleClass)):
Vertex( # type: ignore # this is expected to fail mypy
jvm_view.GaussianVertex, "gaussian", GenericExampleClass(),
GenericExampleClass())
def __create_multi_indexed_samples_generated(
vertices_unwrapped: JavaList, network_samples: JavaObject,
id_to_label: Dict[Tuple[int, ...], str]) -> sample_generator_dict_type:
vertex_samples_multi: Dict = {}
for vertex in vertices_unwrapped:
vertex_label = id_to_label[Vertex._get_python_id(vertex)]
vertex_samples_multi[vertex_label] = defaultdict(list)
sample = Tensor._to_ndarray(network_samples.get(vertex))
__add_sample_to_dict(sample, vertex_samples_multi[vertex_label])
tuple_hierarchy: Dict = {
(vertex_label, tensor_index): values
for vertex_label, tensor_index in vertex_samples_multi.items()
for tensor_index, values in tensor_index.items()
}
return tuple_hierarchy
def test_id_str_of_downstream_vertex_is_higher_than_upstream(
jvm_view: JVMView) -> None:
hyper_params = Vertex(jvm_view.GaussianVertex, 0., 1.)
gaussian = Vertex(jvm_view.GaussianVertex, 0., hyper_params)
hyper_params_id = hyper_params.get_id()
gaussian_id = gaussian.get_id()
assert type(hyper_params_id) == tuple
assert type(gaussian_id) == tuple
assert hyper_params_id < gaussian_id
def __create_multi_indexed_samples(
vertices_unwrapped: JavaList, network_samples: JavaObject,
id_to_label: Dict[Tuple[int, ...], str]) -> sample_types:
vertex_samples_multi: Dict = {}
for vertex in vertices_unwrapped:
vertex_label = id_to_label[Vertex._get_python_id(vertex)]
vertex_samples_multi[vertex_label] = defaultdict(list)
samples_for_vertex = __get_vertex_samples(network_samples, vertex)
for sample in samples_for_vertex:
__add_sample_to_dict(sample, vertex_samples_multi[vertex_label])
tuple_hierarchy: Dict = {
(vertex_label, shape_index): values
for vertex_label, samples in vertex_samples_multi.items()
for shape_index, values in samples.items()
}
return tuple_hierarchy
def sample(net: BayesNet,
sample_from: Iterable[Vertex],
algo: str = 'metropolis',
draws: int = 500,
drop: int = 0,
down_sample_interval: int = 1) -> sample_types:
vertices_unwrapped = k.to_java_object_list(sample_from)
network_samples = algorithms[algo].withDefaultConfig().getPosteriorSamples(
net.unwrap(), vertices_unwrapped,
draws).drop(drop).downSample(down_sample_interval)
vertex_samples = {
Vertex._get_python_id(vertex_unwrapped): list(
map(Tensor._to_ndarray,
network_samples.get(vertex_unwrapped).asList()))
for vertex_unwrapped in vertices_unwrapped
}
return vertex_samples
def test_get_connected_graph(jvm_view: JVMView) -> None:
gaussian = Vertex(jvm_view.GaussianVertex, 0., 1.)
connected_graph = set(gaussian.get_connected_graph())
assert len(connected_graph) == 3
def test_can_pass_array_to_vertex(jvm_view: JVMView) -> None:
gaussian = Vertex(jvm_view.GaussianVertex, "gaussian", [3, 3], Const(0.),
Const(1.))
sample = gaussian.sample()
assert sample.shape == (3, 3)
def test_java_vertex_to_python_vertex_persists_label() -> None:
label = "gaussian"
java_vertex = Gaussian(0., 1., label=label).unwrap()
python_vertex = Vertex._from_java_vertex(java_vertex)
assert python_vertex.get_label() == label
def assert_vertex_value_equals_ndarray(vertex: Vertex, expected_type: Type,
ndarray: numpy_types) -> None:
vertex_value = vertex.get_value()
expected_value = ndarray.astype(expected_type)
assert np.array_equal(vertex_value, expected_value)
assert np.issubdtype(vertex_value.dtype, expected_type)
def test_construct_vertex_with_java_vertex(jvm_view: JVMView) -> None:
java_vertex = Vertex(jvm_view.GaussianVertex, 0., 1.).unwrap()
python_vertex = Vertex(java_vertex)
assert tuple(java_vertex.getId().getValue()) == python_vertex.get_id()
def test_can_pass_ndarray_to_vertex(jvm_view: JVMView) -> None:
gaussian = Vertex(jvm_view.GaussianVertex, np.array([0.1, 0.4]),
np.array([0.4, 0.5]))
sample = gaussian.sample()
assert sample.shape == (2, )
def test_you_can_set_value(vertex: Vertex, expected_type: Type,
value: tensor_arg_types,
assert_vertex_value_equals: Callable) -> None:
vertex.set_value(value)
assert_vertex_value_equals(vertex, expected_type, value)
def test_can_pass_pandas_dataframe_to_vertex(jvm_view: JVMView) -> None:
gaussian = Vertex(jvm_view.GaussianVertex, pd.DataFrame(data=[0.1, 0.4]),
pd.DataFrame(data=[0.1, 0.4]))
sample = gaussian.sample()
assert sample.shape == (2, 1)
def test_construct_vertex_with_java_vertex() -> None:
java_vertex = Gaussian(0., 1.).unwrap()
python_vertex = Vertex._from_java_vertex(java_vertex)
assert tuple(java_vertex.getId().getValue()) == python_vertex.get_id()
def test_can_pass_pandas_series_to_vertex(jvm_view):
gaussian = Vertex(jvm_view.GaussianVertex, pd.Series(data=[0.1, 0.4]),
pd.Series(data=[0.1, 0.4]))
sample = gaussian.sample()
assert sample.shape == (2, )
def __wrap(vertices: JavaMap) -> Dict[str, Vertex]:
return {
k.getUnqualifiedName(): Vertex._from_java_vertex(v)
for k, v in vertices.items()
}
def test_can_pass_array_to_vertex(jvm_view: JVMView) -> None:
gaussian = Vertex(jvm_view.GaussianVertex, [3, 3], 0., 1.)
sample = gaussian.sample()
assert sample.shape == (3, 3)
def get_double_model_output_vertex(self, label: str) -> Vertex:
label_unwrapped = _VertexLabel(label).unwrap()
result = self.unwrap().getDoubleModelOutputVertex(label_unwrapped)
return Vertex._from_java_vertex(result)
def value_for(self, v: Vertex) -> numpy_types:
return Tensor._to_ndarray(self.unwrap().getValueFor(
v.unwrap().getReference()))
