def kriging_eq(simple_model_2_internals):
sp_internal, ori_internal, options = simple_model_2_internals
A_matrix = yield_covariance(InterpInput(sp_internal, ori_internal,
options))
b_vector = yield_b_vector(ori_internal, A_matrix.shape[0])
return A_matrix, b_vector
def test_reduction(self, internals):
sp_internals, ori_internals, options = internals
# Test cov
cov = yield_covariance(InterpInput(sp_internals, ori_internals, options))
print("\n")
print(cov)
np.testing.assert_array_almost_equal(np.asarray(cov), cov_sol, decimal=3)
# Test weights and b vector
b_vec = yield_b_vector(ori_internals, cov.shape[0])
weights = kernel_reduction(cov, b_vec)
print(weights)
weights_gempy_v2 = [6.402e+00, -1.266e+01, 2.255e-15, -2.784e-15, 1.236e+01, 2.829e+01, -6.702e+01, -6.076e+02,
1.637e+03, 1.053e+03, 2.499e+02, -2.266e+03]
np.testing.assert_allclose(np.asarray(weights), weights_gempy_v2, rtol=2)
def test_covariance_cubic_kernel(simple_model_2):
# Cubic kernel
# Euclidean distance
l = np.load(dir_name + '/../solutions/test_kernel_numeric2.npy')
surface_points = simple_model_2[0]
orientations = simple_model_2[1]
options = simple_model_2[2]
tensors_structure = simple_model_2[3]
sp_internals = surface_points_preprocess(surface_points, tensors_structure.number_of_points_per_surface)
ori_internals = orientations_preprocess(orientations)
cov = yield_covariance(InterpInput(sp_internals, ori_internals, options))
print(cov)
print(l)
np.testing.assert_array_almost_equal(np.asarray(cov), l, decimal=3)
gi_res=1,
number_dimensions=2)
_ = np.ones(3)
tensor_structure = TensorsStructure(np.array([3, 2]))
return spi, ori_i, kri, tensor_structure
BackendTensor.change_backend(AvailableBackends.numpy,
use_gpu=True,
pykeops_enabled=True)
surface_points, orientations, options, tensors_structure = simple_model_2()
sp_internals = surface_points_preprocess(
surface_points, tensors_structure.number_of_points_per_surface)
ori_internals = orientations_preprocess(orientations)
kernel_data = cov_vectors_preparation(sp_internals, ori_internals, options)
cov = yield_covariance(sp_internals, ori_internals, options)
cov_grad = _test_covariance_items(kernel_data, options, "cov_grad")
#cov_grad = _test_covariance_items(kernel_data, options, "cov_grad")
# Compile and run cov sum axis 0
#print(cov.sum(axis=0))
# TODO: Compile and run solver
def weights(self, internals):
sp_internals, ori_internals, options = internals
cov = yield_covariance(InterpInput(sp_internals, ori_internals, options))
b_vec = yield_b_vector(ori_internals, cov.shape[0])
weights = kernel_reduction(cov, b_vec)
return weights