def test_plusmin_mstep():
# Set constants
n_seq = 1
T = 2000
# Generate data
plusmin = PlusminModel()
data, hidden = plusmin.generate_dataset(n_seq, T)
n_features = plusmin.x_dim
n_components = plusmin.K
# Fit reference model and initial MSLDS model
refmodel = GaussianHMM(n_components=n_components,
covariance_type='full').fit(data)
warnings.filterwarnings("ignore", category=DeprecationWarning)
# Obtain sufficient statistics from refmodel
rlogprob, rstats = reference_estep(refmodel, data)
means = refmodel.means_
covars = refmodel.covars_
transmat = refmodel.transmat_
populations = refmodel.startprob_
As = []
for i in range(n_components):
As.append(np.zeros((n_features, n_features)))
Qs = refmodel.covars_
bs = refmodel.means_
means = refmodel.means_
covars = refmodel.covars_
# Test AQB solver for MSLDS
solver = MetastableSwitchingLDSSolver(n_components, n_features)
solver.do_mstep(As, Qs, bs, means, covars, rstats)
def test_plusmin_stats():
# Set constants
num_hotstart = 3
n_seq = 1
T = 2000
# Generate data
plusmin = PlusminModel()
data, hidden = plusmin.generate_dataset(n_seq, T)
n_features = plusmin.x_dim
n_components = plusmin.K
# Fit reference model
refmodel = GaussianHMM(n_components=n_components,
covariance_type='full').fit(data)
warnings.filterwarnings("ignore", category=DeprecationWarning)
# Fit initial MSLDS model from reference model
model = MetastableSwitchingLDS(n_components, n_features,
n_hotstart=0)
model.inferrer._sequences = data
model.means_ = refmodel.means_
model.covars_ = refmodel.covars_
model.transmat_ = refmodel.transmat_
model.populations_ = refmodel.startprob_
model.As_ = [np.zeros((n_features, n_features)),
np.zeros((n_features, n_features))]
model.Qs_ = refmodel.covars_
model.bs_ = refmodel.means_
iteration = 0 # Remove this step once hot_start is factored out
logprob, stats = model.inferrer.do_estep()
rlogprob, rstats = reference_estep(refmodel, data)
yield lambda: np.testing.assert_array_almost_equal(stats['post'],
rstats['post'], decimal=3)
yield lambda: np.testing.assert_array_almost_equal(stats['post[1:]'],
rstats['post[1:]'], decimal=3)
yield lambda: np.testing.assert_array_almost_equal(stats['post[:-1]'],
rstats['post[:-1]'], decimal=3)
yield lambda: np.testing.assert_array_almost_equal(stats['obs'],
rstats['obs'], decimal=3)
yield lambda: np.testing.assert_array_almost_equal(stats['obs[1:]'],
rstats['obs[1:]'], decimal=3)
yield lambda: np.testing.assert_array_almost_equal(stats['obs[:-1]'],
rstats['obs[:-1]'], decimal=3)
yield lambda: np.testing.assert_array_almost_equal(stats['obs*obs.T'],
rstats['obs*obs.T'], decimal=3)
yield lambda: np.testing.assert_array_almost_equal(
stats['obs*obs[t-1].T'], rstats['obs*obs[t-1].T'], decimal=3)
yield lambda: np.testing.assert_array_almost_equal(
stats['obs[1:]*obs[1:].T'], rstats['obs[1:]*obs[1:].T'],
decimal=3)
yield lambda: np.testing.assert_array_almost_equal(
stats['obs[:-1]*obs[:-1].T'], rstats['obs[:-1]*obs[:-1].T'],
decimal=3)
yield lambda: np.testing.assert_array_almost_equal(
stats['trans'], rstats['trans'], decimal=1)
def test_plusmin():
# Set constants
n_hotstart = 3
n_em_iter = 3
n_experiments = 1
n_seq = 1
T = 2000
gamma = 512.
# Generate data
plusmin = PlusminModel()
data, hidden = plusmin.generate_dataset(n_seq, T)
n_features = plusmin.x_dim
n_components = plusmin.K
# Train MSLDS
mslds_scores = []
l = MetastableSwitchingLDS(n_components, n_features,
n_hotstart=n_hotstart, n_em_iter=n_em_iter,
n_experiments=n_experiments)
l.fit(data, gamma=gamma)
mslds_score = l.score(data)
print("gamma = %f" % gamma)
print("MSLDS Log-Likelihood = %f" % mslds_score)
print()
# Fit Gaussian HMM for comparison
g = GaussianFusionHMM(plusmin.K, plusmin.x_dim)
g.fit(data)
hmm_score = g.score(data)
print("HMM Log-Likelihood = %f" % hmm_score)
print()
# Plot sample from MSLDS
sim_xs, sim_Ss = l.sample(T, init_state=0, init_obs=plusmin.mus[0])
sim_xs = np.reshape(sim_xs, (n_seq, T, plusmin.x_dim))
plt.close('all')
plt.figure(1)
plt.plot(range(T), data[0], label="Observations")
plt.plot(range(T), sim_xs[0], label='Sampled Observations')
plt.legend()
plt.show()