def test_core_sets(self):
dtrajs = [
np.array([0, 0, 2, 0, 0, 3, 0, 5, 5, 5, 0, 0, 6, 8, 4, 1, 2, 0, 3])
]
expected = [
np.array(
[-1, -1, 2, 2, 2, 3, 3, 5, 5, 5, 5, 5, 6, 6, 4, 4, 2, 2, 3])
]
core_set = np.arange(2, 7)
dts = DiscreteTrajectoryStats(dtrajs)
dts.to_coreset(core_set=core_set)
np.testing.assert_equal(dts.discrete_trajectories, expected)
def test_realistic_random(self):
n_states = 102
n_traj = 10
dtrajs = [
np.random.randint(0, n_states, size=1000) for _ in range(n_traj)
]
core_set = np.random.randint(0, n_states, size=30)
dts = DiscreteTrajectoryStats(dtrajs)
actual = dts.to_coreset(core_set, in_place=False)
def naive(dtrajs, core_set):
import copy
dtrajs = copy.deepcopy(dtrajs)
newdiscretetraj = []
for t, st in enumerate(dtrajs):
oldmicro = -1
newtraj = []
for f, micro in enumerate(st):
newmicro = None
for co in core_set:
if micro == co:
newmicro = micro
oldmicro = micro
break
if newmicro is None and oldmicro is not None:
newtraj.append(oldmicro)
elif newmicro is not None:
newtraj.append(newmicro)
else:
print("hi there")
newtraj.append(-1)
newdiscretetraj.append(np.array(newtraj, dtype=int))
return newdiscretetraj
expected = naive(dtrajs, core_set)
np.testing.assert_equal(actual, expected)
def test_mincount_connectivity(self):
dtrajs = np.zeros(10, dtype=int)
dtrajs[0] = 1
dtrajs[-1] = 1
dts = DiscreteTrajectoryStats(dtrajs)
dts.count_lagged(1, mincount_connectivity=0)
C_mincount0 = dts.count_matrix_largest.todense()
np.testing.assert_equal(C_mincount0, np.array([[7, 1], [1, 0]]))
dts.count_lagged(1, mincount_connectivity=2)
C = np.array(dts.count_matrix_largest.todense())
np.testing.assert_equal(C, np.array([[7]]))
# check that the original count matrix remains unmodified
np.testing.assert_equal(dts.count_matrix().todense(), C_mincount0)
def test_core_sets_5(self):
dtrajs = [np.array([2, 2, 2, 2, 2, 2, 2, 0])]
dts = DiscreteTrajectoryStats(dtrajs)
dts.to_coreset([2])
np.testing.assert_equal(dts.discrete_trajectories,
[np.ones_like(dtrajs[0]) * 2])
def test_core_sets_3(self):
dtrajs = [np.array([2, 0, 1, 1, 2])]
expected = [np.array([2, 2, 1, 1, 2])]
dts = DiscreteTrajectoryStats(dtrajs)
dts.to_coreset(np.arange(1, 3))
np.testing.assert_equal(dts.discrete_trajectories, expected)