def test_chain_graph_single_obs(self):
with self.test_session() as sess:
np.random.seed(1234)
num_observations = 1
num_timesteps = 5
batch_size = 2
state_size = 1
observations, lengths = datasets.create_chain_graph_dataset(
batch_size=batch_size,
num_timesteps=num_timesteps,
state_size=state_size)
out_observations, out_lengths = sess.run([observations, lengths])
self.assertAllEqual([num_observations, num_observations],
out_lengths)
self.assertAllClose([[[1.426677], [-1.789461]]], out_observations)
def test_chain_graph_single_obs(self):
with self.test_session() as sess:
np.random.seed(1234)
num_observations = 1
num_timesteps = 5
batch_size = 2
state_size = 1
observations, lengths = datasets.create_chain_graph_dataset(
batch_size=batch_size,
num_timesteps=num_timesteps,
state_size=state_size)
out_observations, out_lengths = sess.run([observations, lengths])
self.assertAllEqual([num_observations, num_observations], out_lengths)
self.assertAllClose(
[[[1.426677], [-1.789461]]],
out_observations)
def test_chain_graph_state_dims(self):
with self.test_session() as sess:
np.random.seed(1234)
num_observations = 1
num_timesteps = 5
batch_size = 2
state_size = 3
observations, lengths = datasets.create_chain_graph_dataset(
batch_size=batch_size,
num_timesteps=num_timesteps,
state_size=state_size)
out_observations, out_lengths = sess.run([observations, lengths])
self.assertAllEqual([num_observations, num_observations],
out_lengths)
self.assertAllClose([[[1.052287, -4.560759, 3.07988],
[2.008926, 0.495567, 3.488678]]],
out_observations)
def test_chain_graph_state_dims(self):
with self.test_session() as sess:
np.random.seed(1234)
num_observations = 1
num_timesteps = 5
batch_size = 2
state_size = 3
observations, lengths = datasets.create_chain_graph_dataset(
batch_size=batch_size,
num_timesteps=num_timesteps,
state_size=state_size)
out_observations, out_lengths = sess.run([observations, lengths])
self.assertAllEqual([num_observations, num_observations], out_lengths)
self.assertAllClose(
[[[1.052287, -4.560759, 3.07988],
[2.008926, 0.495567, 3.488678]]],
out_observations)
def test_chain_graph_multiple_obs(self):
with self.test_session() as sess:
np.random.seed(1234)
num_observations = 3
num_timesteps = 6
batch_size = 2
state_size = 1
observations, lengths = datasets.create_chain_graph_dataset(
batch_size=batch_size,
num_timesteps=num_timesteps,
steps_per_observation=num_timesteps / num_observations,
state_size=state_size)
out_observations, out_lengths = sess.run([observations, lengths])
self.assertAllEqual([num_observations, num_observations],
out_lengths)
self.assertAllClose(
[[[0.40051451], [1.07405114]], [[1.73932898], [3.16880035]],
[[-1.98377144], [2.82669163]]], out_observations)
def test_chain_graph_fixed_obs(self):
with self.test_session() as sess:
np.random.seed(1234)
num_observations = 3
num_timesteps = 6
batch_size = 2
state_size = 1
observations, lengths = datasets.create_chain_graph_dataset(
batch_size=batch_size,
num_timesteps=num_timesteps,
steps_per_observation=num_timesteps/num_observations,
state_size=state_size,
fixed_observation=4.)
out_observations, out_lengths = sess.run([observations, lengths])
self.assertAllEqual([num_observations, num_observations], out_lengths)
self.assertAllClose(
np.ones([num_observations, batch_size, state_size]) * 4.,
out_observations)
def create_graph():
"""Creates the dataset, model, and bound."""
xs, lengths = datasets.create_chain_graph_dataset(
config.batch_size,
config.num_timesteps,
steps_per_observation=1,
state_size=1,
transition_variance=config.variance,
observation_variance=config.variance)
model = ghmm.TrainableGaussianHMM(config.num_timesteps,
config.proposal_type,
transition_variances=config.variance,
emission_variances=config.variance,
random_seed=config.random_seed)
true_likelihood = tf.reduce_mean(
model.likelihood(tf.squeeze(xs)) / tf.to_float(lengths))
outs = [true_likelihood]
outs.extend(list(create_bound(model, xs, lengths)))
return outs
def test_chain_graph_multiple_obs(self):
with self.test_session() as sess:
np.random.seed(1234)
num_observations = 3
num_timesteps = 6
batch_size = 2
state_size = 1
observations, lengths = datasets.create_chain_graph_dataset(
batch_size=batch_size,
num_timesteps=num_timesteps,
steps_per_observation=num_timesteps/num_observations,
state_size=state_size)
out_observations, out_lengths = sess.run([observations, lengths])
self.assertAllEqual([num_observations, num_observations], out_lengths)
self.assertAllClose(
[[[0.40051451], [1.07405114]],
[[1.73932898], [3.16880035]],
[[-1.98377144], [2.82669163]]],
out_observations)
def create_graph():
"""Creates the training graph."""
global_step = tf.train.get_or_create_global_step()
xs, lengths = datasets.create_chain_graph_dataset(
config.batch_size,
config.num_timesteps,
steps_per_observation=1,
state_size=1,
transition_variance=config.variance,
observation_variance=config.variance)
model = ghmm.TrainableGaussianHMM(config.num_timesteps,
config.proposal_type,
transition_variances=config.variance,
emission_variances=config.variance,
random_seed=config.random_seed)
loss, bound, true_ll, gap = create_losses(model, xs, lengths)
opt = tf.train.AdamOptimizer(config.learning_rate)
grads = opt.compute_gradients(loss, var_list=tf.trainable_variables())
train_op = opt.apply_gradients(grads, global_step=global_step)
return bound, true_ll, gap, train_op, global_step
def create_graph():
"""Creates the dataset, model, and bound."""
xs, lengths = datasets.create_chain_graph_dataset(
config.batch_size,
config.num_timesteps,
steps_per_observation=1,
state_size=1,
transition_variance=config.variance,
observation_variance=config.variance)
model = ghmm.TrainableGaussianHMM(
config.num_timesteps,
config.proposal_type,
transition_variances=config.variance,
emission_variances=config.variance,
random_seed=config.random_seed)
true_likelihood = tf.reduce_mean(
model.likelihood(tf.squeeze(xs)) / tf.to_float(lengths))
outs = [true_likelihood]
outs.extend(list(create_bound(model, xs, lengths)))
return outs
def create_graph():
"""Creates the training graph."""
global_step = tf.train.get_or_create_global_step()
xs, lengths = datasets.create_chain_graph_dataset(
config.batch_size,
config.num_timesteps,
steps_per_observation=1,
state_size=1,
transition_variance=config.variance,
observation_variance=config.variance)
model = ghmm.TrainableGaussianHMM(
config.num_timesteps,
config.proposal_type,
transition_variances=config.variance,
emission_variances=config.variance,
random_seed=config.random_seed)
loss, bound, true_ll, gap = create_losses(model, xs, lengths)
opt = tf.train.AdamOptimizer(config.learning_rate)
grads = opt.compute_gradients(loss, var_list=tf.trainable_variables())
train_op = opt.apply_gradients(grads, global_step=global_step)
return bound, true_ll, gap, train_op, global_step
def test_chain_graph_raises_error_on_wrong_steps_per_observation(self):
with self.assertRaises(ValueError):
datasets.create_chain_graph_dataset(batch_size=4,
num_timesteps=10,
steps_per_observation=9)
def test_chain_graph_raises_error_on_wrong_steps_per_observation(self):
with self.assertRaises(ValueError):
datasets.create_chain_graph_dataset(
batch_size=4,
num_timesteps=10,
steps_per_observation=9)