def test_threshold_match_children_2d():
threshold = gate.ThresholdGate(gate_name="test",
parent="test parent",
x="X",
y="Y",
method="density")
x = np.random.normal(loc=1., scale=1.5, size=1000)
y = np.random.normal(loc=1., scale=1.5, size=1000)
data = pd.DataFrame({"X": x, "Y": y})
threshold.add_child(
gate.ChildThreshold(name="positive",
definition="++,+-",
geom=ThresholdGeom(x_threshold=0.5)))
threshold.add_child(
gate.ChildThreshold(name="negative",
definition="--,-+",
geom=ThresholdGeom(x_threshold=0.5)))
pos = gate.Population(population_name="p1",
parent="root",
definition="++",
geom=ThresholdGeom(x_threshold=0.6),
index=data[data.X >= 0.6].index.values)
neg = gate.Population(population_name="p2",
parent="root",
definition="--,-+",
geom=ThresholdGeom(x_threshold=0.6),
index=data[data.X < 0.6].index.values)
pops = threshold._match_to_children([neg, pos])
pos = [p for p in pops if p.definition == "++"][0]
assert pos.population_name == "positive"
neg = [p for p in pops if p.definition == "--,-+"][0]
assert neg.population_name == "negative"
def test_threshold_match_children_1d():
threshold = gate.ThresholdGate(gate_name="test",
parent="test parent",
x="X",
method="density")
data = np.random.normal(loc=1., scale=1.5, size=1000)
threshold.add_child(
gate.ChildThreshold(name="positive",
definition="+",
geom=ThresholdGeom(x_threshold=0.5)))
threshold.add_child(
gate.ChildThreshold(name="negative",
definition="-",
geom=ThresholdGeom(x_threshold=0.5)))
pos = gate.Population(population_name="p1",
parent="root",
definition="+",
geom=ThresholdGeom(x_threshold=0.6),
index=data[np.where(data >= 0.6)])
neg = gate.Population(population_name="p2",
parent="root",
definition="-",
geom=ThresholdGeom(x_threshold=0.6),
index=data[np.where(data >= 0.6)])
pops = threshold._match_to_children([neg, pos])
pos = [p for p in pops if p.definition == "+"][0]
assert pos.population_name == "positive"
neg = [p for p in pops if p.definition == "-"][0]
assert neg.population_name == "negative"
def test_threshold_add_child_invalid_1d(d):
threshold = gate.ThresholdGate(gate_name="test",
parent="test parent",
method="manual",
x="X")
child = gate.ChildThreshold(name="test child",
definition=d,
geom=ThresholdGeom(x="X",
x_threshold=0.56,
y_threshold=0.75))
with pytest.raises(AssertionError) as err:
threshold.add_child(child)
assert str(
err.value) == "Invalid child definition, should be either '+' or '-'"
def test_threshold_predict_2d():
data, _ = make_blobs(n_samples=3000,
n_features=2,
centers=[(1., 1.), (1., 5.), (5., 0.2)],
random_state=42)
data = pd.DataFrame({"X": data[:, 0], "Y": data[:, 1]})
threshold = gate.ThresholdGate(gate_name="test",
parent="test parent",
x="X",
y="Y",
method="density")
threshold.fit(data=data)
new_data, _ = make_blobs(n_samples=3000,
n_features=2,
centers=[(1., 1.), (5., 0.2)],
random_state=42)
new_data = pd.DataFrame({"X": new_data[:, 0], "Y": new_data[:, 1]})
pops = threshold.predict(new_data)
assert len(pops) == 4
assert all([isinstance(p, gate.Population) for p in pops])
assert all([isinstance(p.geom, ThresholdGeom) for p in pops])
assert all([p.geom.x == threshold.x for p in pops])
assert all([p.geom.y == threshold.y for p in pops])
assert all(p.geom.transform_x == threshold.transformations.get("x")
for p in pops)
assert all(p.geom.transform_y == threshold.transformations.get("y")
for p in pops)
assert all(i in [p.definition for p in pops]
for i in ["++", "--", "-+", "+-"])
neg_idx = new_data[
(new_data.X < threshold.children[0].geom.x_threshold)
& (new_data.Y < threshold.children[0].geom.y_threshold)].index.values
pos_idx = new_data[
(new_data.X >= threshold.children[0].geom.x_threshold)
& (new_data.Y >= threshold.children[0].geom.y_threshold)].index.values
negpos_idx = new_data[
(new_data.X < threshold.children[0].geom.x_threshold)
& (new_data.Y >= threshold.children[0].geom.y_threshold)].index.values
posneg_idx = new_data[
(new_data.X >= threshold.children[0].geom.x_threshold)
& (new_data.Y < threshold.children[0].geom.y_threshold)].index.values
pos_pop = [p for p in pops if p.definition == "++"][0]
neg_pop = [p for p in pops if p.definition == "--"][0]
posneg_pop = [p for p in pops if p.definition == "+-"][0]
negpos_pop = [p for p in pops if p.definition == "-+"][0]
assert np.array_equal(neg_pop.index, neg_idx)
assert np.array_equal(pos_pop.index, pos_idx)
assert np.array_equal(negpos_pop.index, negpos_idx)
assert np.array_equal(posneg_pop.index, posneg_idx)
def test_threshold_fit_1d():
np.random.seed(42)
n1 = np.random.normal(loc=0.2, scale=1, size=500)
n2 = np.random.normal(loc=2.5, scale=0.2, size=250)
n3 = np.random.normal(loc=6.5, scale=0.5, size=500)
data = pd.DataFrame({"X": np.hstack([n1, n2, n3])})
threshold = gate.ThresholdGate(gate_name="test",
parent="test parent",
x="X",
method="density")
threshold.fit(data=data)
assert len(threshold.children) == 2
assert threshold.children[0].geom.x_threshold == threshold.children[
1].geom.x_threshold
assert round(threshold.children[0].geom.x_threshold) == 4
assert all(
[i in [c.definition for c in threshold.children] for i in ["+", "-"]])
def test_threshold_add_child():
threshold = gate.ThresholdGate(gate_name="test",
parent="test parent",
x="X",
y="Y",
method="manual",
transformations={"x": "logicle"})
child = gate.ChildThreshold(name="test child",
definition="++",
geom=ThresholdGeom(x_threshold=0.56,
y_threshold=0.75))
threshold.add_child(child)
assert len(threshold.children)
assert threshold.children[0].geom.x == threshold.x
assert threshold.children[0].geom.y == threshold.y
assert threshold.children[0].geom.transform_x == "logicle"
assert not threshold.children[0].geom.transform_y
def test_threshold_fit_2d():
data, labels = make_blobs(n_samples=3000,
n_features=2,
centers=[(1., 1.), (1., 5.), (5., 0.2)],
random_state=42)
data = pd.DataFrame({"X": data[:, 0], "Y": data[:, 1]})
threshold = gate.ThresholdGate(gate_name="test",
parent="test parent",
x="X",
y="Y",
method="density")
threshold.fit(data)
assert len(threshold.children) == 4
assert len(set([c.geom.x_threshold for c in threshold.children])) == 1
assert len(set([c.geom.y_threshold for c in threshold.children])) == 1
assert all([
i in [c.definition for c in threshold.children]
for i in ["++", "--", "+-", "-+"]
])
assert 2 < threshold.children[0].geom.x_threshold < 4
assert 2 < threshold.children[0].geom.y_threshold < 4
def test_threshold_fit_predict_2d():
data, _ = make_blobs(n_samples=4000,
n_features=2,
centers=[(1., 1.), (1., 7.), (7., 2.), (7., 6.2)],
random_state=42)
data = pd.DataFrame({"X": data[:, 0], "Y": data[:, 1]})
threshold = gate.ThresholdGate(gate_name="test",
parent="test parent",
x="X",
y="Y",
method="density")
threshold.fit(data)
threshold.label_children({
"++": "Top left",
"--": "Other",
"-+": "Other",
"+-": "Other"
})
data, _ = make_blobs(n_samples=3000,
n_features=2,
centers=[(1., 1.), (1., 7.), (7., 6.2)],
random_state=42)
data = pd.DataFrame({"X": data[:, 0], "Y": data[:, 1]})
pops = threshold.fit_predict(data=data)
assert len(pops) == 2
assert all([isinstance(p, gate.Population) for p in pops])
assert all([isinstance(p.geom, ThresholdGeom) for p in pops])
assert all([p.geom.x == threshold.x for p in pops])
assert all([p.geom.y == threshold.y for p in pops])
assert all(p.geom.transform_x == threshold.transformations.get("x")
for p in pops)
assert all(p.geom.transform_y == threshold.transformations.get("y")
for p in pops)
top_left = [p for p in pops if p.population_name == "Top left"][0]
other = [p for p in pops if p.population_name == "Other"][0]
assert top_left.definition == "++"
assert {"+-", "-+", "--"} == set(other.definition.split(","))
assert len(top_left.index) < len(other.index)
assert len(top_left.index) > 900
assert len(other.index) > 1900
def test_threshold_fit_predict_1d():
n1 = np.random.normal(loc=0.2, scale=1, size=500)
n2 = np.random.normal(loc=2.5, scale=0.2, size=250)
n3 = np.random.normal(loc=6.5, scale=0.5, size=500)
data = pd.DataFrame({"X": np.hstack([n1, n2, n3])})
threshold = gate.ThresholdGate(gate_name="test",
parent="test parent",
x="X",
method="density")
threshold.fit(data=data)
threshold.label_children({"+": "Positive", "-": "Negative"})
new_data = pd.DataFrame({
"X":
np.hstack([
np.random.normal(loc=0.2, scale=1, size=200),
np.random.normal(loc=6.5, scale=0.5, size=1000)
])
})
pops = threshold.fit_predict(new_data)
assert len(pops) == 2
assert all([isinstance(p, gate.Population) for p in pops])
assert all([isinstance(p.geom, ThresholdGeom) for p in pops])
assert all([p.geom.x == threshold.x for p in pops])
assert all([p.geom.y == threshold.y for p in pops])
assert all(p.geom.transform_x == threshold.transformations.get("x")
for p in pops)
assert all(p.geom.transform_y == threshold.transformations.get("y")
for p in pops)
assert all(i in [p.definition for p in pops] for i in ["+", "-"])
pos_pop = [p for p in pops if p.definition == "+"][0]
assert pos_pop.population_name == "Positive"
neg_pop = [p for p in pops if p.definition == "-"][0]
assert neg_pop.population_name == "Negative"
assert len(pos_pop.index) > len(neg_pop.index)
assert len(pos_pop.index) > 800
assert len(neg_pop.index) < 300
def test_threshold_predict_1d():
n1 = np.random.normal(loc=0.2, scale=1, size=500)
n2 = np.random.normal(loc=2.5, scale=0.2, size=250)
n3 = np.random.normal(loc=6.5, scale=0.5, size=500)
data = pd.DataFrame({"X": np.hstack([n1, n2, n3])})
threshold = gate.ThresholdGate(gate_name="test",
parent="test parent",
x="X",
method="density")
threshold.fit(data=data)
new_data = pd.DataFrame({
"X":
np.hstack([
np.random.normal(loc=0.2, scale=1, size=500),
np.random.normal(loc=6.5, scale=0.5, size=500)
])
})
pops = threshold.predict(new_data)
assert len(pops) == 2
assert all([isinstance(p, gate.Population) for p in pops])
assert all([isinstance(p.geom, ThresholdGeom) for p in pops])
assert all([p.geom.x == threshold.x for p in pops])
assert all([p.geom.y == threshold.y for p in pops])
assert all(p.geom.transform_x == threshold.transformations.get("x")
for p in pops)
assert all(p.geom.transform_y == threshold.transformations.get("y")
for p in pops)
assert all(i in [p.definition for p in pops] for i in ["+", "-"])
neg_idx = new_data[
new_data.X < threshold.children[0].geom.x_threshold].index.values
pos_idx = new_data[
new_data.X >= threshold.children[0].geom.x_threshold].index.values
pos_pop = [p for p in pops if p.definition == "+"][0]
neg_pop = [p for p in pops if p.definition == "-"][0]
assert np.array_equal(neg_pop.index, neg_idx)
assert np.array_equal(pos_pop.index, pos_idx)