def test_Histogram_discrete_inference(self):
data = np.array([1, 1, 2, 3, 3, 3]).reshape(-1, 1)
ds_context = Context([MetaType.DISCRETE])
ds_context.add_domains(data)
hist = create_histogram_leaf(data, ds_context, [0], alpha=False)
prob = np.exp(log_likelihood(hist, data))
self.assertAlmostEqual(float(prob[0]), 2 / 6)
self.assertAlmostEqual(float(prob[1]), 2 / 6)
self.assertAlmostEqual(float(prob[2]), 1 / 6)
self.assertAlmostEqual(float(prob[3]), 3 / 6)
self.assertAlmostEqual(float(prob[4]), 3 / 6)
self.assertAlmostEqual(float(prob[5]), 3 / 6)
data = np.array([1, 1, 2, 3, 3, 3]).reshape(-1, 1)
ds_context = Context([MetaType.DISCRETE])
ds_context.add_domains(data)
hist = create_histogram_leaf(data, ds_context, [0], alpha=True)
# print(np.var(data.shape[0]))
prob = np.exp(log_likelihood(hist, data))
self.assertAlmostEqual(float(prob[0]), 3 / 9)
self.assertAlmostEqual(float(prob[1]), 3 / 9)
self.assertAlmostEqual(float(prob[2]), 2 / 9)
self.assertAlmostEqual(float(prob[3]), 4 / 9)
self.assertAlmostEqual(float(prob[4]), 4 / 9)
self.assertAlmostEqual(float(prob[5]), 4 / 9)
def create_piecewise_leaf(data, ds_context, scope, isotonic=False, prior_weight=0.1, hist_source="numpy"):
assert len(scope) == 1, "scope of univariate Piecewise for more than one variable?"
assert data.shape[1] == 1, "data has more than one feature?"
idx = scope[0]
meta_type = ds_context.meta_types[idx]
hist = create_histogram_leaf(data, ds_context, scope, alpha=False, hist_source=hist_source)
densities = hist.densities
bins = hist.breaks
repr_points = hist.bin_repr_points
if meta_type == MetaType.REAL:
EPS = 1e-8
if len(densities) > 1:
def pairwise(iterable):
"s -> (s0,s1), (s1,s2), (s2, s3), ..."
a, b = itertools.tee(iterable)
next(b, None)
return zip(a, b)
x = [bins[0] - EPS] + [b0 + (b1 - b0) / 2 for (b0, b1) in pairwise(bins)] + [bins[-1] + EPS]
else:
assert len(bins) == 2
x = [bins[0] - EPS] + [(bins[0] + (bins[1] - bins[0]) / 2)] + [bins[-1] + EPS]
elif meta_type == MetaType.DISCRETE:
tail_width = 1
x = [b for b in bins[:-1]]
x = [x[0] - tail_width] + x + [x[-1] + tail_width]
else:
raise Exception("Invalid statistical type: " + meta_type)
y = [0.0] + [d for d in densities] + [0.0]
assert len(densities) == len(bins) - 1
assert len(x) == len(y), (len(x), len(y))
x, y = np.array(x), np.array(y)
if isotonic:
x, y = isotonic_unimodal_regression_R(x, y)
auc = np.trapz(y, x)
y = y / auc
node = PiecewiseLinear(x.tolist(), y.tolist(), repr_points, scope=scope)
if prior_weight is None:
return node
uniform_data = np.zeros_like(data)
uniform_data[:] = np.nan
uniform_hist = create_histogram_leaf(uniform_data, ds_context, scope, alpha=False)
return prior_weight * uniform_hist + (1 - prior_weight) * node
def test_Histogram_expectations(self):
data = np.random.randn(20000).reshape(-1, 1)
ds_context = Context(meta_types=[MetaType.REAL])
ds_context.add_domains(data)
hl = create_histogram_leaf(data, ds_context, scope=[0])
expectation = Expectation(hl, set([0]))
self.assertAlmostEqual(np.mean(data[:, 0]), expectation[0, 0], 3)
data = np.random.randint(0, high=100, size=20000).reshape(-1, 1)
ds_context = Context(meta_types=[MetaType.DISCRETE])
ds_context.add_domains(data)
hl = create_histogram_leaf(data, ds_context, scope=[0])
expectation = Expectation(hl, set([0]))
self.assertAlmostEqual(np.mean(data[:, 0]), expectation[0, 0], 3)
def test_histogram_to_str_and_back(self):
data = np.array([1, 1, 2, 3, 3, 3]).reshape(-1, 1)
ds_context = Context([MetaType.DISCRETE])
ds_context.add_domains(data)
hist = create_histogram_leaf(data, ds_context, [0], alpha=False)
self.check_obj_and_reconstruction(hist)
def test_histogram_leaf(self):
data = np.array([1, 1, 2, 3, 3, 3]).reshape(-1, 1)
ds_context = Context([MetaType.DISCRETE])
ds_context.add_domains(data)
hist = create_histogram_leaf(data, ds_context, [0], alpha=False)
self.assertTrue(
np.array_equal(mpe(hist, np.array([[np.nan]])), np.array([[3]])),
"mpe should be 3")
def test_valid_histogram(self):
np.random.seed(17)
data = [1] + [5]*20 + [7] + [10]*50 + [20] + [30]*10
data = np.array(data).reshape((-1, 1))
ds_context = Context([MetaType.REAL])
ds_context.add_domains(data)
hist = create_histogram_leaf(data, ds_context, [0], alpha=False, hist_source="kde")
self.assertGreater(len(hist.bin_repr_points), 1)
def create_spmn_leaf(data, ds_context, scope):
assert len(
scope
) == 1, "scope of univariate histogram for more than one variable?"
assert data.shape[1] == 1, "data has more than one feature?"
# data = data[~np.isnan(data)]
idx = scope[0]
meta_type = ds_context.meta_types[idx]
if meta_type == MetaType.UTILITY:
hist = create_histogram_leaf(data, ds_context, scope)
return Utility(hist.breaks,
hist.densities,
hist.bin_repr_points,
scope=idx)
else:
return create_histogram_leaf(data, ds_context, scope)
def test_mixture_gaussians(self):
np.random.seed(17)
data = np.random.normal(10, 1, size=200).tolist() + np.random.normal(30, 1, size=200).tolist()
data = np.array(data).reshape((-1, 1))
ds_context = Context([MetaType.REAL])
ds_context.add_domains(data)
hist = create_histogram_leaf(data, ds_context, [0], alpha=False, hist_source="kde")
x = np.linspace(0, 60, 1000).tolist() + data[:, 0].tolist()
x = np.sort(x)
from scipy.stats import norm
y = 0.5 * norm.pdf(x, 10, 1) + 0.5 * norm.pdf(x, 30, 1)
ye = likelihood(hist, x.reshape((-1, 1)))
error = np.sum(np.abs(ye[:, 0] - y))
# print(error)
self.assertLessEqual(error, 7)
def learn_leaf_from_context(data, ds_context, scope):
"""
Wrapper function to infer leaf type from the context object
:param data: np.array: the data slice
:param ds_context: Context: the context oobject for the data/spn
:param scope: List: the scope of the variables
:return: a correct leaf
"""
assert len(scope) == 1, "scope for more than one variable?"
idx = scope[0]
conditional_type = ds_context.parametric_types[idx]
assert issubclass(conditional_type, Leaf), 'no instance of leaf '
if issubclass(conditional_type, Parametric):
return create_parametric_leaf(data, ds_context, scope)
if issubclass(conditional_type, Conditional):
return create_conditional_leaf(data, ds_context, scope)
if issubclass(conditional_type, Histogram):
return create_histogram_leaf(data, ds_context, scope)
if issubclass(conditional_type, PiecewiseLinear):
return create_piecewise_leaf(data, ds_context, scope)
raise Exception('No fitting leaf type found')
def create_leaf(data, ds_context, scope):
#return create_piecewise_leaf(data, ds_context, scope, isotonic=False, prior_weight=0.01)
return create_histogram_leaf(data, ds_context, scope, alpha=0.005)
if __name__ == '__main__':
add_histogram_inference_support()
np.random.seed(17)
data = np.random.normal(10, 0.01, size=2000).tolist() + np.random.normal(
30, 10, size=2000).tolist()
data = np.array(data).reshape((-1, 10))
data[data < 0] = 0
data = (data * 1).astype(int)
ds_context = Context(meta_types=[MetaType.DISCRETE] * data.shape[1])
ds_context.add_domains(data)
data[:, 0] = 0
data[:, 1] = 1
spn = learn(data, ds_context)
spn = create_histogram_leaf(data[:, 0].reshape((-1, 1)), ds_context, [0], alpha=False, hist_source="kde") * \
create_histogram_leaf(data[:, 1].reshape((-1, 1)), ds_context, [1], alpha=False, hist_source="kde")
spn = 0.3 * create_histogram_leaf(data[:, 0].reshape((-1, 1)), ds_context, [0], alpha=False, hist_source="kde") + \
0.7 * create_histogram_leaf(data[:, 0].reshape((-1, 1)), ds_context, [0], alpha=False, hist_source="kde")
py_ll = log_likelihood(spn, data)
tf_graph, placeholder = spn_to_tf_graph(spn, data)
log_tf_out = eval_tf(tf_graph, placeholder, data)
print("results are similar for Log TF and Python?",
np.all(np.isclose(py_ll, log_tf_out)))
