def test_insert_lag_time(self):
lags = [1, 3, 5]
its = timescales_msm(self.dtraj2, lags=lags, errors='bayes', nsamples=10, show_progress=False)
new_lags = np.concatenate((lags, [2, 4]+list(range(6, 9))), axis=0)
its.lags = new_lags
np.testing.assert_equal(its._lags, new_lags)
its.estimate(self.dtraj2)
# compare with a one shot estimation
its_one_shot = timescales_msm(self.dtraj2, lags=new_lags, errors='bayes', nsamples=10, show_progress=False)
np.testing.assert_equal(its.timescales, its_one_shot.timescales)
self.assertEqual([m.lag for m in its.models],
[m.lag for m in its_one_shot.models])
# estimate with different data to trigger re-estimation
from pyemma.util.testing_tools import MockLoggingHandler
log_handler = MockLoggingHandler()
its.logger.addHandler(log_handler)
extended_new_lags = new_lags.tolist()
extended_new_lags.append(20)
its.estimate(self.dtraj4_2, lags=extended_new_lags)
np.testing.assert_equal(its.models[0].dtrajs_full[0], self.dtraj4_2)
self.assertIn("estimating from new data", log_handler.messages['warning'][0])
# remove a lag time and ensure the corresponding model is removed too
new_lags = new_lags[:-3]
new_lags_len = len(new_lags)
its.lags = new_lags
np.testing.assert_equal(its.lags, new_lags)
assert len(its.models) == new_lags_len
assert len(its.timescales) == new_lags_len
assert len(its.sample_mean) == new_lags_len
def test_insert_remove_lag_time(self):
# test insert and removal at the same time
lags = [1, 3, 5]
its = timescales_msm(self.dtraj4_2, lags=lags, errors='bayes', nsamples=10, show_progress=False)
new_lags = lags + [6, 7, 8]
new_lags = new_lags[2:]
new_lags += [21, 22]
# omit the first lag
new_lags = new_lags[1:]
its.estimate(self.dtraj4_2, lags=new_lags)
its_one_shot = timescales_msm(self.dtraj4_2, lags=new_lags)
np.testing.assert_allclose(its.timescales, its_one_shot.timescales)
def test_2_parallel(self):
t2 = timescales(self.P2)[1]
lags = [1, 2, 3, 4, 5]
its = timescales_msm([self.dtraj2], lags=lags, n_jobs=2)
est = its.timescales[0]
assert (np.alltrue(est < t2 + 2.0))
assert (np.alltrue(est > t2 - 2.0))
def test_2_parallel(self):
t2 = timescales(self.P2)[1]
lags = [1, 2, 3, 4, 5]
its = timescales_msm([self.dtraj2], lags=lags, n_jobs=2)
est = its.timescales[0]
np.testing.assert_array_less(est, t2 + 2.0)
np.testing.assert_array_less(t2 - 2.0, est)
def test_no_return_estimators(self):
lags = [1, 2, 3, 10, 20]
nstates = 10
nits = 3
its = timescales_msm(dtrajs=np.random.randint(0, nstates, size=1000), lags=lags,
only_timescales=True, nits=nits)
with self.assertRaises(RuntimeError):
its.estimators
with self.assertRaises(RuntimeError):
its.models
assert isinstance(its.timescales, np.ndarray)
assert its.timescales.shape == (len(lags), nstates - 1 if nstates == nits else nits)
def test_errors(self):
dtraj_disconnected = [0, 0, 0, 0, -1, 1, 1, 1, 1]
with self.assertRaises(RuntimeError) as e:
timescales_msm(dtraj_disconnected, lags=[1, 2, 3, 4, 5])
self.assertIn('negative row index', e.exception.args[0])
def test_errors(self):
dtraj_disconnected = [-2] * 10
with self.assertRaises(RuntimeError) as e:
timescales_msm(dtraj_disconnected, lags=[1, 2, 3, 4, 5])
self.assertIn('elements < -1', e.exception.args[0])
