def setup(self):
"""
Setup function will create a fixed point set and parameter
settings for testing different aspects of this library.
"""
self.X = generate_test_grid_2d(40)
self.Y = gerber(self.X)
self.norm_x = {}
scaler = sklearn.preprocessing.MinMaxScaler()
self.norm_x["feature"] = scaler.fit_transform(np.atleast_2d(self.X))
self.norm_x["zscore"] = sklearn.preprocessing.scale(self.X,
axis=0,
with_mean=True,
with_std=True,
copy=True)
self.norm_x["none"] = self.X
# Methods covered here:
# __init__
# build
# __set_data
self.graph = ngl.EmptyRegionGraph(max_neighbors=10)
self.ct = topopy.ContourTree(graph=self.graph, debug=False)
self.ct.build(self.X, self.Y)
def setup(self):
"""
Setup function will create a fixed point set and parameter
settings for testing different aspects of this library.
"""
self.X = generate_test_grid_2d(40)
self.Y = gerber(self.X)
self.graph = ngl.EmptyRegionGraph(max_neighbors=10)
self.norm_x = {}
scaler = sklearn.preprocessing.MinMaxScaler()
self.norm_x["feature"] = scaler.fit_transform(np.atleast_2d(self.X))
self.norm_x["zscore"] = sklearn.preprocessing.scale(self.X,
axis=0,
with_mean=True,
with_std=True,
copy=True)
self.norm_x["none"] = self.X
# Methods covered here:
# __init__
# build
# __set_data
self.test_object = topopy.MorseComplex(debug=False, graph=self.graph)
self.test_object.build(self.X, self.Y)
gold_path = os.path.join("topopy", "tests", "mc_gold.json")
with open(gold_path, "r") as data_file:
gold_json = data_file.read()
gold_json = json.loads(gold_json)
self.gold = gold_json
def test_neighbors(self):
""" Tests the neighbors function in both settings, that is where
an index is supplied and when it is not. This does not use
an input neighborhood graph, thus NGL must prune the
complete graph in this case.
"""
self.setup()
graph_rep = nglpy.EmptyRegionGraph(max_neighbors=self.max_neighbors,
beta=self.beta,
relaxed=self.relaxed,
p=self.p,
discrete_steps=self.discrete_steps)
graph_rep.build(self.points)
expected_graph = {
0: (1, ),
1: (0, 2, 4),
2: (1, 3),
3: (2, ),
4: (1, )
}
for i in range(len(self.points)):
expected = list(expected_graph[i])
actual = sorted(graph_rep.neighbors(i))
msg = "\nNode {} Connectivity:".format(i)
msg += "\n\texpected: {}\n\tactual: {} ".format(expected, actual)
self.assertEqual(expected, actual, msg)
self.assertEqual(graph_rep.neighbors(), expected_graph)
def test_empty(self):
""" Tests handling of empty data, we just want to make sure nothing
breaks terribly.
"""
self.setup()
graph_rep = nglpy.EmptyRegionGraph(max_neighbors=self.max_neighbors,
beta=self.beta,
relaxed=True,
p=self.p,
discrete_steps=self.discrete_steps)
graph_rep.build([])
expected_graph = {}
self.assertEqual(graph_rep.neighbors(), expected_graph)
def setup(self):
"""
Setup function will create a fixed point set and parameter
settings for testing different aspects of this library.
"""
self.X = generate_test_grid_2d(40)
self.Y = gerber(self.X)
self.graph = ngl.EmptyRegionGraph(max_neighbors=10)
self.norm_x = {}
scaler = sklearn.preprocessing.MinMaxScaler()
self.norm_x["feature"] = scaler.fit_transform(np.atleast_2d(self.X))
self.norm_x["zscore"] = sklearn.preprocessing.scale(
self.X, axis=0, with_mean=True, with_std=True, copy=True
)
self.norm_x["none"] = self.X
def __init__(
self,
graph=None,
gradient="steepest",
normalization=None,
aggregator=None,
debug=False,
):
super(TopologicalObject, self).__init__()
self.reset()
if graph is None:
graph = ngl.EmptyRegionGraph()
self.graph = graph
self.gradient = gradient
self.normalization = normalization
self.debug = debug
self.aggregator = aggregator
def get_cbsg(df, beta=0):
graph = nx.Graph()
graph.add_nodes_from(df.iterrows())
point_set = np.array(df[["x", "y"]])
point_set = point_set + 0.0001
# aGraph = nglpy.Graph(point_set, "beta skeleton", 9, beta)
aGraph = nglpy.EmptyRegionGraph(max_neighbors=9, relaxed=False, beta=beta)
aGraph.build(point_set)
d = aGraph.neighbors()
for key, value in d.items():
for v in value:
graph.add_edge(
key, v, weight=euclidean(df.loc[key, ["x", "y"]], df.loc[v, ["x", "y"]])
)
return graph
def test_aggregation(self):
# Since the default function can change, here we will only test
# that the correct number of each array is reported
X = np.ones((11, 2))
X[10] = [0, 0]
Y = np.array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 100])
warnings.filterwarnings("ignore")
x, y = topopy.TopologicalObject.aggregate_duplicates(X, Y)
self.assertEqual(
2,
len(x),
"aggregate_duplicates should return a list of "
"unique items in X.",
)
self.assertEqual(
2,
len(y),
"aggregate_duplicates should return the aggregated "
"values of Y for each unique item in X.",
)
# Next, we will test each of the string aggregation function
# names on scalar Y values
x, y = topopy.TopologicalObject.aggregate_duplicates(X, Y, "min")
self.assertListEqual(x.tolist(), [[0, 0], [1, 1]])
self.assertListEqual(y.tolist(), [100, 0])
x, y = topopy.TopologicalObject.aggregate_duplicates(X, Y, "max")
self.assertListEqual(x.tolist(), [[0, 0], [1, 1]])
self.assertListEqual(y.tolist(), [100, 9])
x, y = topopy.TopologicalObject.aggregate_duplicates(X, Y, "mean")
self.assertListEqual(x.tolist(), [[0, 0], [1, 1]])
self.assertListEqual(y.tolist(), [100, 4.5])
x, y = topopy.TopologicalObject.aggregate_duplicates(X, Y, "average")
self.assertListEqual(x.tolist(), [[0, 0], [1, 1]])
self.assertListEqual(y.tolist(), [100, 4.5])
x, y = topopy.TopologicalObject.aggregate_duplicates(X, Y, "median")
self.assertListEqual(x.tolist(), [[0, 0], [1, 1]])
self.assertListEqual(y.tolist(), [100, 4.5])
# Next, we will test each of the string aggregation function
# names on vector Y values
Y = np.array(
[
[0, 9],
[1, 8],
[2, 7],
[3, 6],
[4, 5],
[5, 4],
[6, 3],
[7, 2],
[8, 1],
[9, 0],
[100, 0],
]
)
x, y = topopy.TopologicalObject.aggregate_duplicates(X, Y, "min")
self.assertListEqual(x.tolist(), [[0, 0], [1, 1]])
self.assertListEqual(y.tolist(), [[100, 0], [0, 0]])
x, y = topopy.TopologicalObject.aggregate_duplicates(X, Y, "max")
self.assertListEqual(x.tolist(), [[0, 0], [1, 1]])
self.assertListEqual(y.tolist(), [[100, 0], [9, 9]])
x, y = topopy.TopologicalObject.aggregate_duplicates(X, Y, "mean")
self.assertListEqual(x.tolist(), [[0, 0], [1, 1]])
self.assertListEqual(y.tolist(), [[100, 0], [4.5, 4.5]])
x, y = topopy.TopologicalObject.aggregate_duplicates(X, Y, "median")
self.assertListEqual(x.tolist(), [[0, 0], [1, 1]])
self.assertListEqual(y.tolist(), [[100, 0], [4.5, 4.5]])
x, y = topopy.TopologicalObject.aggregate_duplicates(X, Y, "first")
self.assertListEqual(x.tolist(), [[0, 0], [1, 1]])
self.assertListEqual(y.tolist(), [[100, 0], [0, 9]])
x, y = topopy.TopologicalObject.aggregate_duplicates(X, Y, "last")
self.assertListEqual(x.tolist(), [[0, 0], [1, 1]])
self.assertListEqual(y.tolist(), [[100, 0], [9, 0]])
# Testing custom callable aggregator
x, y = topopy.TopologicalObject.aggregate_duplicates(
X, Y, lambda x: x[0]
)
self.assertListEqual(x.tolist(), [[0, 0], [1, 1]])
self.assertListEqual(y.tolist(), [[100, 0], [0, 9]])
warnings.filterwarnings("always")
# Testing an invalid aggregator
with warnings.catch_warnings(record=True) as w:
x, y = topopy.TopologicalObject.aggregate_duplicates(
X, Y, "invalid"
)
self.assertTrue(issubclass(w[-1].category, UserWarning))
self.assertEqual(
'Aggregator "invalid" not understood. Skipping sample '
'aggregation.',
str(w[-1].message),
)
self.assertListEqual(x.tolist(), X.tolist())
self.assertListEqual(y.tolist(), Y.tolist())
# Testing aggregator on non-duplicate data
X = np.array([[0, 0], [0, 1]])
Y = np.array([0, 1])
x, y = topopy.TopologicalObject.aggregate_duplicates(X, Y)
self.assertListEqual(x.tolist(), X.tolist())
self.assertListEqual(y.tolist(), Y.tolist())
warnings.filterwarnings("ignore")
# Testing use of the aggregator in the check_duplicates function
X = np.ones((11, 2))
X[10] = [0, 0]
Y = np.array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 100])
to = topopy.TopologicalObject()
self.assertRaises(ValueError, to.build, **{"X": X, "Y": Y})
graph = ngl.EmptyRegionGraph(max_neighbors=10)
to = topopy.TopologicalObject(aggregator="mean", graph=graph)
to.build(X, Y)
warnings.filterwarnings("always")