def test_exp_biases(temporal_graph):
rw = TemporalRandomWalk(temporal_graph)
times = np.array([1, 2, 3])
t_0 = 1
expected = np.exp(t_0 - times) / sum(np.exp(t_0 - times))
biases = rw._exp_biases(times, t_0, decay=True)
assert np.allclose(biases, expected)
def test_not_progressing_enough(temporal_graph):
rw = TemporalRandomWalk(temporal_graph)
cw_size = 5 # no valid temporal walks of this size
with pytest.raises(RuntimeError, match=r".* discarded .*"):
rw.run(num_cw=1, cw_size=cw_size, max_walk_length=cw_size, seed=None)
def test_exp_biases_extreme(temporal_graph):
rw = TemporalRandomWalk(temporal_graph)
large_times = [100000, 100001]
biases = rw._exp_biases(large_times, t_0=0, decay=True)
assert sum(biases) == pytest.approx(1)
small_times = [0.000001, 0.000002]
biases = rw._exp_biases(small_times, t_0=0, decay=True)
assert sum(biases) == pytest.approx(1)
def test_temporal_walks(temporal_graph):
"""
valid time respecting walks (node -[time]-> node):
1 -[2]-> 2 -[10]-> 4
2 -[10]-> 4 -[12]-> 6
3 -[2]-> 2 -[10]-> 4
5 -[4]-> 4 -[12]-> 6
1 -[2]-> 2 -[10]-> 4 -[12]-> 6
3 -[2]-> 2 -[10]-> 4 -[12] -> 6
"""
expected = {(1, 2, 4), (2, 4, 6), (3, 2, 4), (5, 4, 6), (1, 2, 4, 6),
(3, 2, 4, 6)}
rw = TemporalRandomWalk(temporal_graph)
num_cw = 20 # how many walks to be sure we're getting valid temporal walks
for walk in rw.run(num_cw=num_cw, cw_size=3, max_walk_length=4, seed=None):
assert tuple(walk) in expected
def test_cw_size_and_walk_length(temporal_graph, cw_size):
rw = TemporalRandomWalk(temporal_graph)
num_cw = 5
max_walk_length = 3
def run():
return rw.run(num_cw=num_cw, cw_size=cw_size, max_walk_length=max_walk_length)
if cw_size < 2:
with pytest.raises(ValueError, match=r".* context window size .*"):
run()
elif max_walk_length < cw_size:
with pytest.raises(ValueError, match=r".* maximum walk length .*"):
run()
else:
walks = run()
num_cw_obtained = sum([len(walk) - cw_size + 1 for walk in walks])
assert num_cw == num_cw_obtained
assert max(map(len, walks)) <= max_walk_length
def test_init_parameters(temporal_graph):
num_cw = 5
cw_size = 3
max_walk_length = 3
seed = 0
rw = TemporalRandomWalk(
temporal_graph, cw_size=cw_size, max_walk_length=max_walk_length, seed=seed
)
rw_no_params = TemporalRandomWalk(temporal_graph)
run_1 = rw.run(num_cw=num_cw)
run_2 = rw_no_params.run(
num_cw=num_cw, cw_size=cw_size, max_walk_length=max_walk_length, seed=seed
)
assert np.array_equal(run_1, run_2)
def test_init_parameters(temporal_graph):
num_cw = 5
cw_size = 3
max_walk_length = 3
seed = 0
rw = TemporalRandomWalk(temporal_graph,
cw_size=cw_size,
max_walk_length=max_walk_length,
seed=seed)
rw_no_params = TemporalRandomWalk(temporal_graph)
assert rw.run(num_cw=num_cw) == rw_no_params.run(
num_cw=num_cw,
cw_size=cw_size,
max_walk_length=max_walk_length,
seed=seed)