def test_inverse_transform_spd(self):
point = spd.SPDMatrices(3).random_uniform(10)
transformer = ToTangentSpace(geometry=spd.SPDMetricLogEuclidean(3))
X = transformer.fit_transform(X=point)
result = transformer.inverse_transform(X)
expected = point
self.assertAllClose(expected, result, atol=1e-4)
transformer = ToTangentSpace(geometry=spd.SPDMetricAffine(3))
X = transformer.fit_transform(X=point)
result = transformer.inverse_transform(X)
expected = point
self.assertAllClose(expected, result, atol=1e-4)
def test_inverse_transform_hyperbolic(self):
points = self.hyperbolic.random_uniform(10)
transformer = ToTangentSpace(geometry=self.hyperbolic.metric)
X = transformer.fit_transform(X=points)
result = transformer.inverse_transform(X)
expected = points
self.assertAllClose(expected, result)
def test_inverse_transform_so(self):
point = self.so_matrix.random_uniform(10)
transformer = ToTangentSpace(
geometry=self.so_matrix.bi_invariant_metric)
X = transformer.transform(X=point, base_point=self.so_matrix.identity)
result = transformer.inverse_transform(
X, base_point=self.so_matrix.identity)
expected = point
self.assertAllClose(expected, result)
def test_inverse_transform_no_fit_sphere(self):
point = self.sphere.random_uniform(3)
base_point = point[0]
point = point[1:]
transformer = ToTangentSpace(geometry=self.sphere)
X = transformer.transform(point, base_point=base_point)
result = transformer.inverse_transform(X, base_point=base_point)
expected = point
self.assertAllClose(expected, result)
break
features = np.concatenate((features, data[0].reshape((1, image_size)).numpy()))
labels.append(data[1])
labels = np.array(labels)
pca = PCA(n_components=2)
features = pca.fit_transform(features)
scaler = StandardScaler()
features = scaler.fit_transform(features) * 3
features = np.concatenate((features, np.zeros_like(features[:, 0:1])), axis=1)
sphere = Hypersphere(dim=2)
transformer = ToTangentSpace(sphere)
base_point = np.array([0, 0, 1])
s_points = transformer.inverse_transform(features, base_point)
classifier = KNearestNeighborsClassifier()
classifier.fit(s_points, labels)
f_labels = []
def loss_f(x):
label = classifier.predict(x[np.newaxis, :])
f_labels.append(label)
return label == 0
fig = plt.figure(figsize=(10, 10))
ax = fig.add_subplot(111, projection='3d')
plot_sphere = Sphere()