def evaluate(self, test_data):
if len(self.model) == 0:
raise RuntimeError('PolyACO must be trained before evaluation')
inside = np.empty((len(self.planes), len(self.class_indices), test_data.shape[0]), dtype=bool)
for i, plane in enumerate(self.model):
plane_data = np.take(test_data, list(self.planes[i]), axis=1)
for j, poly in enumerate(plane):
inside[i, j, :] = is_points_inside_cuda(plane_data, polygon_to_array(poly))
aggregated_scores = np.mean(inside, axis=0)
predictions = np.empty((test_data.shape[0]), dtype=int)
max_elements = np.argmax(aggregated_scores, axis=0)
for i in range(test_data.shape[0]):
predictions[i] = self.class_indices[max_elements[i]]
return predictions
def evaluate(self, test_data):
if len(self.model) == 0:
raise RuntimeError('PolyACO must be trained before evaluation')
inside = np.empty(
(len(self.planes), len(self.class_indices), test_data.shape[0]),
dtype=bool)
for i, plane in enumerate(self.model):
plane_data = np.take(test_data, list(self.planes[i]), axis=1)
for j, poly in enumerate(plane):
inside[i,
j, :] = is_points_inside_cuda(plane_data,
polygon_to_array(poly))
aggregated_scores = np.mean(inside, axis=0)
predictions = np.empty((test_data.shape[0]), dtype=int)
max_elements = np.argmax(aggregated_scores, axis=0)
for i in range(test_data.shape[0]):
predictions[i] = self.class_indices[max_elements[i]]
return predictions
def cost_function_gpu(points, edges):
is_inside = ray_cast.is_points_inside_cuda(points, edges)
score = np.sum(np.logical_xor(is_inside, points[:, 2]))
return score / points.shape[0]
def test_function_returns_1_true_rest_false(self):
data = np.array([[0, 0], [1, 1], [2, 2], [3, 3]])
result = rc.is_points_inside_cuda(data, self.polygon)
self.assertEqual(np.sum(result), 1)
self.assertEqual(np.sum(np.invert(result)), 3)
def test_function_returns_array_of_size_equal_to_number_of_input_points(self):
data = np.array([[0, 0], [1, 1], [2, 2], [3, 3]])
result = rc.is_points_inside_cuda(data, self.polygon)
self.assertEqual(result.shape[0], data.shape[0])
def cost_function_gpu(points, edges):
is_inside = ray_cast.is_points_inside_cuda(points, edges)
score = np.sum(np.logical_xor(is_inside, points[:, 2]))
return score / points.shape[0]
def test_function_returns_1_true_rest_false(self):
data = np.array([[0, 0], [1, 1], [2, 2], [3, 3]])
result = rc.is_points_inside_cuda(data, self.polygon)
self.assertEqual(np.sum(result), 1)
self.assertEqual(np.sum(np.invert(result)), 3)
def test_function_returns_array_of_size_equal_to_number_of_input_points(
self):
data = np.array([[0, 0], [1, 1], [2, 2], [3, 3]])
result = rc.is_points_inside_cuda(data, self.polygon)
self.assertEqual(result.shape[0], data.shape[0])