def test_bellman_ford_reference(self):
Edges = np.array([[1, 4], [3, 1], [1, 3], [0, 1], [0, 2], [3, 2],
[1, 2], [4, 3]])
w = np.array([2, 1, 2, -1, 4, 5, 3, -3], dtype=float)
G = sparse.coo_matrix((w, (Edges[:, 0], Edges[:, 1])))
# distance
# seed 0 : [ 0. -1. 2. -2. 1.]
# seed 1 : [inf 0. 3. -1. 2.]
# seed 2 : [inf inf 0. inf inf]
# seed 3 : [inf 1. 4. 0. 3.]
# seed 4 : [inf -2. 1. -3. 0.]
distances_FROM_seed = np.array([[0., -1., 2., -2., 1.],
[np.inf, 0., 3., -1., 2.],
[np.inf, np.inf, 0., np.inf, np.inf],
[np.inf, 1., 4., 0., 3.],
[np.inf, -2., 1., -3., 0.]])
distances_FROM_seed[np.where(
distances_FROM_seed == np.inf)] = max_value(G.dtype)
for seed in range(5):
distance, nearest = reference_bellman_ford(G.T, [seed])
assert_equal(distance, distances_FROM_seed[seed])
distance, nearest = bellman_ford(G.T, [seed])
assert_equal(distance, distances_FROM_seed[seed])
# seeds [0,1,2,3,4]
# distance to closest: [-2. -1. 0. 0. -3.]
# closest: [3 3 2 3 3]
distance_TO_closest = np.array([-2., -1., 0., 0., -3.])
ref_closest = np.array([3, 3, 2, 3, 3])
distance, closest = reference_bellman_ford(G, [0, 1, 2, 3, 4])
assert_equal(distance, distance_TO_closest)
assert_equal(closest, ref_closest)
distance, closest = bellman_ford(G, [0, 1, 2, 3, 4])
assert_equal(distance, distance_TO_closest)
assert_equal(closest, ref_closest)
def test_bellman_ford(self):
np.random.seed(1643502758)
for G in self.cases:
G.data = np.random.rand(G.nnz)
N = G.shape[0]
for n_seeds in [int(N / 20), int(N / 10), N - 2, N]:
if n_seeds > G.shape[0] or n_seeds < 1:
continue
seeds = np.random.permutation(N)[:n_seeds]
D_expected, S_expected = reference_bellman_ford(G, seeds)
D_result, S_result = bellman_ford(G, seeds)
assert_equal(D_result, D_expected)
assert_equal(S_result, S_expected)
def test_bellman_ford_reference(self):
Edges = np.array([[1, 4], [3, 1], [1, 3], [0, 1], [0, 2], [3, 2],
[1, 2], [4, 3]])
w = np.array([2, 1, 2, 1, 4, 5, 3, 1], dtype=float)
G = sparse.coo_matrix((w, (Edges[:, 0], Edges[:, 1])))
distances_FROM_seed = np.array([[0., 1., 4., 3., 3.],
[np.inf, 0., 3., 2., 2.],
[np.inf, np.inf, 0., np.inf, np.inf],
[np.inf, 1., 4., 0., 3.],
[np.inf, 2., 5., 1., 0.]])
for seed in range(5):
distance, nearest = bellman_ford_reference(G, [seed])
assert_equal(distance, distances_FROM_seed[seed])
distance, nearest = bellman_ford(G, [seed])
assert_equal(distance, distances_FROM_seed[seed])
def test_bellman_ford(self):
numpy.random.seed(0)
for G in self.cases:
G.data = rand(G.nnz)
N = G.shape[0]
for n_seeds in [int(N/20), int(N/10), N-2, N]:
if n_seeds > G.shape[0] or n_seeds < 1:
continue
seeds = numpy.random.permutation(N)[:n_seeds]
D_expected, S_expected = reference_bellman_ford(G, seeds)
D_result, S_result = bellman_ford(G, seeds)
assert_equal(D_result, D_expected)
assert_equal(S_result, S_expected)
if __name__ == '__main__':
Edges = np.array([[1, 4],
[3, 1],
[1, 3],
[0, 1],
[0, 2],
[3, 2],
[1, 2],
[4, 3]])
w = np.array([2, 1, 2, 1, 4, 5, 3, 1], dtype=float)
A = sparse.coo_matrix((w, (Edges[:, 0], Edges[:, 1])))
c = np.array([0,1,2,3,4])
print('\nreference--')
for cc in c:
d, m = bellman_ford_reference(A, [cc])
print(d, m)
print('\npyamg--')
from pyamg.graph import bellman_ford
for cc in c:
d, m = bellman_ford(A, [cc])
print(d, m)
print('\ncsgraph.bellman_ford')
from scipy.sparse import csgraph
for cc in c:
d, p = csgraph.bellman_ford(A, directed=True, indices=[cc], return_predecessors=True)
print(d.ravel(), p.ravel())