def test_binary_grid_unaries():
# test handling on unaries for binary grid CRFs
for ds in binary:
X, Y = ds(n_samples=1)
x, y = X[0], Y[0]
for inference_method in get_installed():
# dai is to expensive
crf = GridCRF(inference_method=inference_method)
crf.initialize(X, Y)
w_unaries_only = np.zeros(7)
w_unaries_only[:4] = np.eye(2).ravel()
# test that inference with unaries only is the
# same as argmax
inf_unaries = crf.inference(x, w_unaries_only)
assert_array_equal(
inf_unaries, np.argmax(x, axis=2),
"Wrong unary inference for %s" % inference_method)
try:
assert (np.mean(inf_unaries == y) > 0.5)
except:
print(ds)
# check that the right thing happens on noise-free data
X, Y = ds(n_samples=1, noise=0)
inf_unaries = crf.inference(X[0], w_unaries_only)
assert_array_equal(inf_unaries, Y[0],
"Wrong unary result for %s" % inference_method)
def test_binary_blocks():
X, Y = generate_blocks(n_samples=10)
crf = GridCRF()
clf = StructuredPerceptron(model=crf, max_iter=40)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
def test_multinomial_blocks_frankwolfe():
X, Y = generate_blocks_multinomial(n_samples=10, noise=0.5, seed=0)
crf = GridCRF(inference_method='qpbo')
clf = FrankWolfeSSVM(model=crf, C=1, max_iter=50, verbose=3)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
def test_blocks_multinomial_crf():
X, Y = generate_blocks_multinomial(n_samples=1, size_x=9, seed=0)
x, y = X[0], Y[0]
w = np.array([
1.,
0.,
0., # unaryA
0.,
1.,
0.,
0.,
0.,
1.,
.4, # pairwise
-.3,
.3,
-.5,
-.1,
.3
])
for inference_method in get_installed():
crf = GridCRF(inference_method=inference_method)
crf.initialize(X, Y)
y_hat = crf.inference(x, w)
assert_array_equal(y, y_hat)
def test_multinomial_blocks():
X, Y = generate_blocks_multinomial(n_samples=10, noise=0.3, seed=0)
crf = GridCRF(n_states=X.shape[-1])
clf = StructuredPerceptron(model=crf, max_iter=10)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
def test_switch_to_ad3():
# test if switching between qpbo and ad3 works
if not get_installed(['qpbo']) or not get_installed(['ad3']):
return
X, Y = generate_blocks_multinomial(n_samples=5, noise=1.5, seed=0)
crf = GridCRF(n_states=3, inference_method='qpbo')
ssvm = NSlackSSVM(crf, max_iter=10000)
ssvm_with_switch = NSlackSSVM(crf, max_iter=10000, switch_to=('ad3'))
ssvm.fit(X, Y)
ssvm_with_switch.fit(X, Y)
assert_equal(ssvm_with_switch.model.inference_method, 'ad3')
# we check that the dual is higher with ad3 inference
# as it might use the relaxation, that is pretty much guraranteed
assert_greater(ssvm_with_switch.objective_curve_[-1],
ssvm.objective_curve_[-1])
# test that convergence also results in switch
ssvm_with_switch = NSlackSSVM(crf,
max_iter=10000,
switch_to=('ad3'),
tol=10)
ssvm_with_switch.fit(X, Y)
assert_equal(ssvm_with_switch.model.inference_method, 'ad3')
def test_binary_grid_unaries():
# test handling on unaries for binary grid CRFs
for ds in binary:
X, Y = ds(n_samples=1)
x, y = X[0], Y[0]
for inference_method in get_installed():
#NOTE: ad3+ fails because it requires a different data structure
if inference_method == 'ad3+': continue
crf = GridCRF(inference_method=inference_method)
crf.initialize(X, Y)
w_unaries_only = np.zeros(7)
w_unaries_only[:4] = np.eye(2).ravel()
# test that inference with unaries only is the
# same as argmax
inf_unaries = crf.inference(x, w_unaries_only)
assert_array_equal(inf_unaries, np.argmax(x, axis=2),
"Wrong unary inference for %s"
% inference_method)
assert(np.mean(inf_unaries == y) > 0.5)
# check that the right thing happens on noise-free data
X, Y = ds(n_samples=1, noise=0)
inf_unaries = crf.inference(X[0], w_unaries_only)
assert_array_equal(inf_unaries, Y[0],
"Wrong unary result for %s"
% inference_method)
def test_one_slack_constraint_caching():
# testing cutting plane ssvm on easy multinomial dataset
X, Y = generate_blocks_multinomial(n_samples=10, noise=0.5, seed=0,
size_x=9)
n_labels = len(np.unique(Y))
exact_inference = get_installed([('ad3', {'branch_and_bound': True}), "lp"])[0]
crf = GridCRF(n_states=n_labels, inference_method=exact_inference)
clf = OneSlackSSVM(model=crf, max_iter=150, C=1,
check_constraints=True, break_on_bad=True,
inference_cache=50, inactive_window=0)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
assert_equal(len(clf.inference_cache_), len(X))
# there should be 13 constraints, which are less than the 94 iterations
# that are done
# check that we didn't change the behavior of how we construct the cache
constraints_per_sample = [len(cache) for cache in clf.inference_cache_]
if exact_inference == "lp":
assert_equal(len(clf.inference_cache_[0]), 18)
assert_equal(np.max(constraints_per_sample), 18)
assert_equal(np.min(constraints_per_sample), 18)
else:
assert_equal(len(clf.inference_cache_[0]), 13)
assert_equal(np.max(constraints_per_sample), 20)
assert_equal(np.min(constraints_per_sample), 11)
def test_one_slack_constraint_caching():
#testing cutting plane ssvm on easy multinomial dataset
X, Y = generate_blocks_multinomial(n_samples=10,
noise=0.5,
seed=0,
size_x=9)
n_labels = len(np.unique(Y))
crf = GridCRF(n_states=n_labels, inference_method='lp')
clf = OneSlackSSVM(model=crf,
max_iter=150,
C=1,
check_constraints=True,
break_on_bad=True,
inference_cache=50,
inactive_window=0,
verbose=10)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
assert_equal(len(clf.inference_cache_), len(X))
# there should be 11 constraints, which are less than the 94 iterations
# that are done
assert_equal(len(clf.inference_cache_[0]), 11)
# check that we didn't change the behavior of how we construct the cache
constraints_per_sample = [len(cache) for cache in clf.inference_cache_]
assert_equal(np.max(constraints_per_sample), 19)
assert_equal(np.min(constraints_per_sample), 11)
def define_model(self):
if self.models == 'GridCRF':
from pystruct.models import GridCRF
self.crf = GridCRF(
neighborhood=self.models_parameters['neighborhood'],
inference_method=self.models_parameters['inference']
)
if self.models == 'GraphCRF':
self.prepare_data_to_graph_crf()
logging.info('Class weight: ' + str([self.total_label_one_avg,1-self.total_label_one_avg]))
from pystruct.models import GraphCRF
self.crf = GraphCRF(
inference_method=self.models_parameters['inference'],
directed = self.models_parameters['directed'],
class_weight = [self.total_label_one_avg,1-self.total_label_one_avg]
# class_weight=[0.01, 0.99]
)
if self.models == 'EdgeFeatureGraphCRF':
self.prepare_data_to_edge_feature_graph_crf()
logging.info('Class weight: ' + str([self.total_label_one_avg, 1 - self.total_label_one_avg]))
from pystruct.models import EdgeFeatureGraphCRF
self.crf = EdgeFeatureGraphCRF(
inference_method=self.models_parameters['inference'],
# directed=self.models_parameters['directed'],
class_weight=[self.total_label_one_avg, 1 - self.total_label_one_avg]
# class_weight=[0.01, 0.99]
)
return
def test_multinomial_checker_cutting_plane():
X, Y = generate_checker_multinomial(n_samples=10, noise=.1)
n_labels = len(np.unique(Y))
crf = GridCRF(n_states=n_labels, inference_method=inference_method)
clf = NSlackSSVM(model=crf, max_iter=20, C=100000, check_constraints=True)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
def test_multinomial_checker_subgradient():
X, Y = generate_checker_multinomial(n_samples=10, noise=0.4)
n_labels = len(np.unique(Y))
crf = GridCRF(n_states=n_labels, inference_method=inference_method)
clf = SubgradientSSVM(model=crf, max_iter=50)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
def test_binary_blocks_batches_n_slack():
#testing cutting plane ssvm on easy binary dataset
X, Y = generate_blocks(n_samples=5)
crf = GridCRF(inference_method=inference_method)
clf = NSlackSSVM(model=crf, max_iter=20, batch_size=1, C=100)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
def test_binary_blocks():
#testing subgradient ssvm on easy binary dataset
X, Y = generate_blocks(n_samples=5)
crf = GridCRF(inference_method=inference_method)
clf = SubgradientSSVM(model=crf)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
def test_binary_blocks_perceptron_online():
#testing subgradient ssvm on easy binary dataset
X, Y = generate_blocks(n_samples=10)
inference_method = get_installed(['qpbo', 'ad3', 'lp'])[0]
crf = GridCRF(inference_method=inference_method)
clf = StructuredPerceptron(model=crf, max_iter=20)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
def test_binary_blocks_cutting_plane():
#testing cutting plane ssvm on easy binary dataset
X, Y = generate_blocks(n_samples=5)
crf = GridCRF(inference_method=inference_method)
clf = NSlackSSVM(model=crf, max_iter=20, C=100,
check_constraints=True, break_on_bad=False)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
def test_multinomial_blocks_subgradient():
#testing cutting plane ssvm on easy multinomial dataset
X, Y = generate_blocks_multinomial(n_samples=10, noise=0.6, seed=1)
n_labels = len(np.unique(Y))
crf = GridCRF(n_states=n_labels, inference_method=inference_method)
clf = SubgradientSSVM(model=crf, max_iter=50)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
def test_multinomial_blocks_cutting_plane():
#testing cutting plane ssvm on easy multinomial dataset
X, Y = generate_blocks_multinomial(n_samples=40, noise=0.5, seed=0)
n_labels = len(np.unique(Y))
crf = GridCRF(n_states=n_labels, inference_method=inference_method)
clf = NSlackSSVM(model=crf, max_iter=100, C=100, check_constraints=False,
batch_size=1)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
def test_binary_ssvm_attractive_potentials():
# test that submodular SSVM can learn the block dataset
X, Y = generate_blocks(n_samples=10)
crf = GridCRF(inference_method=inference_method)
submodular_clf = NSlackSSVM(model=crf, max_iter=200, C=100,
check_constraints=True,
negativity_constraint=[5])
submodular_clf.fit(X, Y)
Y_pred = submodular_clf.predict(X)
assert_array_equal(Y, Y_pred)
assert_true(submodular_clf.w[5] < 0)
def test_multinomial_blocks_one_slack():
#testing cutting plane ssvm on easy multinomial dataset
X, Y = generate_blocks_multinomial(n_samples=10, noise=0.5, seed=0)
n_labels = len(np.unique(Y))
crf = GridCRF(n_states=n_labels, inference_method=inference_method)
clf = OneSlackSSVM(model=crf, max_iter=150, C=1,
check_constraints=True, break_on_bad=True, tol=.1,
inference_cache=50)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
def test_multinomial_blocks_frankwolfe():
X, Y = generate_blocks_multinomial(n_samples=50, noise=0.5, seed=0)
crf = GridCRF(inference_method='qpbo')
clf = FrankWolfeSSVM(model=crf,
C=1,
line_search=True,
batch_mode=False,
check_dual_every=500)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
def test_binary_blocks_crf_n8_lp():
X, Y = generate_blocks(n_samples=1, noise=1)
x, y = X[0], Y[0]
w = np.array([1, 0, # unary
0, 1,
1, # pairwise
-1.4, 1])
crf = GridCRF(neighborhood=8)
crf.initialize(X, Y)
y_hat = crf.inference(x, w)
assert_array_equal(y, y_hat)
def test_binary_blocks_crf():
X, Y = generate_blocks(n_samples=1)
x, y = X[0], Y[0]
w = np.array([1, 0, # unary
0, 1,
0, # pairwise
-4, 0])
for inference_method in get_installed(['dai', 'qpbo', 'lp', 'ad3', 'ogm']):
crf = GridCRF(inference_method=inference_method)
crf.initialize(X, Y)
y_hat = crf.inference(x, w)
assert_array_equal(y, y_hat)
def test_averaged():
# Under a lot of noise, averaging helps. This fails with less noise.
X, Y = generate_blocks_multinomial(n_samples=15, noise=3, seed=0)
X_train, Y_train = X[:10], Y[:10]
X_test, Y_test = X[10:], Y[10:]
crf = GridCRF()
clf = StructuredPerceptron(model=crf, max_iter=3)
clf.fit(X_train, Y_train)
no_avg_test = clf.score(X_test, Y_test)
clf.set_params(average=True)
clf.fit(X_train, Y_train)
avg_test = clf.score(X_test, Y_test)
assert_greater(avg_test, no_avg_test)
def test_binary_crf_exhaustive():
# tests qpbo inference against brute force
# on random data / weights
np.random.seed(0)
for i in range(10):
x = np.random.uniform(-1, 1, size=(3, 2))
x = np.dstack([-x, np.zeros_like(x)]).copy()
crf = GridCRF(n_features=2, n_states=2)
w = np.random.uniform(-1, 1, size=7)
# check map inference
y_hat = crf.inference(x, w)
y_ex = exhaustive_inference(crf, x, w)
assert_array_equal(y_hat, y_ex)
def test_binary_blocks():
#testing subgradient ssvm on easy binary dataset
X, Y = generate_blocks(n_samples=5)
crf = GridCRF(inference_method=inference_method)
clf = SubgradientSSVM(model=crf,
C=100,
learning_rate=1,
decay_exponent=1,
momentum=0,
decay_t0=10)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
def test_loss_augmentation():
X, Y = generate_blocks(n_samples=1)
x, y = X[0], Y[0]
w = np.array([1, 0, # unary
0, 1,
0, # pairwise
-4, 0])
crf = GridCRF()
crf.initialize(X, Y)
y_hat, energy = crf.loss_augmented_inference(x, y, w, return_energy=True)
assert_almost_equal(energy + crf.loss(y, y_hat),
-np.dot(w, crf.psi(x, y_hat)))
def test_energy_lp():
# make sure that energy as computed by ssvm is the same as by lp
np.random.seed(0)
found_fractional = False
for inference_method in get_installed(["lp", "ad3"]):
crf = GridCRF(n_states=3, n_features=4,
inference_method=inference_method)
while not found_fractional:
x = np.random.normal(size=(2, 2, 4))
w = np.random.uniform(size=crf.size_psi)
inf_res, energy_lp = crf.inference(x, w, relaxed=True,
return_energy=True)
assert_almost_equal(energy_lp,
-np.dot(w, crf.psi(x, inf_res)))
found_fractional = np.any(np.max(inf_res[0], axis=-1) != 1)
def test_blocks_multinomial_crf():
X, Y = generate_blocks_multinomial(n_samples=1, size_x=9, seed=0)
x, y = X[0], Y[0]
w = np.array([1., 0., 0., # unaryA
0., 1., 0.,
0., 0., 1.,
.4, # pairwise
-.3, .3,
-.5, -.1, .3])
for inference_method in get_installed():
#NOTE: ad3+ fails because it requires a different data structure
if inference_method == 'ad3+': continue
crf = GridCRF(inference_method=inference_method)
crf.initialize(X, Y)
y_hat = crf.inference(x, w)
assert_array_equal(y, y_hat)
def test_max_product_binary_blocks():
X, Y = generate_blocks(n_samples=1)
x, y = X[0], Y[0]
w = np.array([
1,
0, # unary
0,
1,
0, # pairwise
-4,
0
])
crf = GridCRF(inference_method='max-product')
crf.initialize(X, Y)
y_hat = crf.inference(x, w)
assert_array_equal(y, y_hat)
