def test_multinomial_blocks_subgradient_batch():
#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=100, batch_size=-1)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
clf2 = SubgradientSSVM(model=crf, max_iter=100, batch_size=len(X))
clf2.fit(X, Y)
Y_pred2 = clf2.predict(X)
assert_array_equal(Y, Y_pred2)
def test_latent_node_boxes_standard_latent():
# learn the "easy" 2x2 boxes dataset.
# a 2x2 box is placed randomly in a 4x4 grid
# we add a latent variable for each 2x2 patch
# that should make the model fairly simple
X, Y = make_simple_2x2(seed=1, n_samples=40)
latent_crf = LatentNodeCRF(n_labels=2, n_hidden_states=2, n_features=1)
one_slack = OneSlackSSVM(latent_crf)
n_slack = NSlackSSVM(latent_crf)
subgradient = SubgradientSSVM(latent_crf, max_iter=100)
for base_svm in [one_slack, n_slack, subgradient]:
base_svm.C = 10
latent_svm = LatentSSVM(base_svm, latent_iter=10)
G = [make_grid_edges(x) for x in X]
# make edges for hidden states:
edges = make_edges_2x2()
G = [np.vstack([make_grid_edges(x), edges]) for x in X]
# reshape / flatten x and y
X_flat = [x.reshape(-1, 1) for x in X]
Y_flat = [y.ravel() for y in Y]
X_ = zip(X_flat, G, [2 * 2 for x in X_flat])
latent_svm.fit(X_[:20], Y_flat[:20])
assert_array_equal(latent_svm.predict(X_[:20]), Y_flat[:20])
assert_equal(latent_svm.score(X_[:20], Y_flat[:20]), 1)
# test that score is not always 1
assert_true(.98 < latent_svm.score(X_[20:], Y_flat[20:]) < 1)
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 fresh_train(self, x, y, iterations=10):
self.model = EdgeFeatureGraphCRF(inference_method="max-product")
self.learner = SubgradientSSVM(
model=self.model,
max_iter=iterations,
logger=SaveLogger(model_file.format(self.userId + "-learner")))
self.learner.fit(x, y, warm_start=False)
self.save()
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 fresh_train(self, x, y, iterations=10):
self.model = ChainCRF(inference_method="max-product")
self.learner = SubgradientSSVM(
model=self.model,
max_iter=iterations,
logger=SaveLogger(
MODEL_PATH_TEMPLATE.format(self.userId + "-learner")),
show_loss_every=50)
self.learner.fit(x, y, warm_start=False)
self.save()
def test_blobs_2d_subgradient():
# make two gaussian blobs
X, Y = make_blobs(n_samples=80, centers=3, random_state=42)
# we have to add a constant 1 feature by hand :-/
X = np.hstack([X, np.ones((X.shape[0], 1))])
X_train, X_test, Y_train, Y_test = X[:40], X[40:], Y[:40], Y[40:]
pbl = MultiClassClf(n_features=3, n_classes=3)
svm = SubgradientSSVM(pbl, C=1000)
svm.fit(X_train, Y_train)
assert_array_equal(Y_test, np.hstack(svm.predict(X_test)))
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_multinomial_blocks_subgradient():
#testing cutting plane ssvm on easy multinomial dataset
X, Y = generate_blocks_multinomial(n_samples=10, noise=0.3, 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,
C=10,
momentum=.98,
learning_rate=0.001)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(Y, Y_pred)
def test_subgradient_svm_as_crf_pickling():
iris = load_iris()
X, y = iris.data, iris.target
X_ = [(np.atleast_2d(x), np.empty((0, 2), dtype=np.int)) for x in X]
Y = y.reshape(-1, 1)
X_train, X_test, y_train, y_test = train_test_split(X_, Y, random_state=1)
_, file_name = mkstemp()
pbl = GraphCRF(n_features=4, n_states=3, inference_method='unary')
logger = SaveLogger(file_name)
svm = SubgradientSSVM(pbl, logger=logger, max_iter=100)
svm.fit(X_train, y_train)
assert_less(.97, svm.score(X_test, y_test))
assert_less(.97, logger.load().score(X_test, y_test))
def test_latent_node_boxes_standard_latent_features():
# learn the "easy" 2x2 boxes dataset.
# we make it even easier now by adding features that encode the correct
# latent state. This basically tests that the features are actually used
X, Y = make_simple_2x2(seed=1, n_samples=20, n_flips=6)
latent_crf = LatentNodeCRF(n_labels=2,
n_hidden_states=2,
n_features=1,
latent_node_features=True)
one_slack = OneSlackSSVM(latent_crf)
n_slack = NSlackSSVM(latent_crf)
subgradient = SubgradientSSVM(latent_crf,
max_iter=100,
learning_rate=0.01,
momentum=0)
for base_svm in [one_slack, n_slack, subgradient]:
base_svm.C = 10
latent_svm = LatentSSVM(base_svm, latent_iter=10)
G = [make_grid_edges(x) for x in X]
# make edges for hidden states:
edges = make_edges_2x2()
G = [np.vstack([make_grid_edges(x), edges]) for x in X]
# reshape / flatten x and y
X_flat = [x.reshape(-1, 1) for x in X]
# augment X with the features for hidden units
X_flat = [
np.vstack([x, y[::2, ::2].reshape(-1, 1)])
for x, y in zip(X_flat, Y)
]
Y_flat = [y.ravel() for y in Y]
X_ = zip(X_flat, G, [2 * 2 for x in X_flat])
latent_svm.fit(X_[:10], Y_flat[:10])
assert_array_equal(latent_svm.predict(X_[:10]), Y_flat[:10])
assert_equal(latent_svm.score(X_[:10], Y_flat[:10]), 1)
# we actually become prefect ^^
assert_true(.98 < latent_svm.score(X_[10:], Y_flat[10:]) <= 1)
def test_with_crosses_base_svms():
# very simple dataset. k-means init is perfect
n_labels = 2
crf = LatentGridCRF(n_labels=n_labels, n_states_per_label=[1, 2])
one_slack = OneSlackSSVM(crf, inference_cache=50)
n_slack = NSlackSSVM(crf)
subgradient = SubgradientSSVM(crf,
max_iter=400,
learning_rate=.01,
decay_exponent=0,
decay_t0=10)
X, Y = generate_crosses(n_samples=10, noise=5, n_crosses=1, total_size=8)
for base_ssvm in [one_slack, n_slack, subgradient]:
base_ssvm.C = 100.
clf = LatentSSVM(base_ssvm=base_ssvm)
clf.fit(X, Y)
Y_pred = clf.predict(X)
assert_array_equal(np.array(Y_pred), Y)
assert_equal(clf.score(X, Y), 1)
def test_ssvm_objectives():
# test that the algorithms provide consistent objective curves.
# this is not that strong a test now but at least makes sure that
# the objective function is called.
X, Y = generate_blocks_multinomial(n_samples=10, noise=1.5, seed=0)
n_labels = len(np.unique(Y))
crf = GridCRF(n_states=n_labels, inference_method=inference_method)
# once for n-slack
clf = NSlackSSVM(model=crf, max_iter=5, C=1, tol=.1)
clf.fit(X, Y)
primal_objective = objective_primal(clf.model, clf.w, X, Y, clf.C)
assert_almost_equal(clf.primal_objective_curve_[-1], primal_objective)
# once for one-slack
clf = OneSlackSSVM(model=crf, max_iter=5, C=1, tol=.1)
clf.fit(X, Y)
primal_objective = objective_primal(clf.model,
clf.w,
X,
Y,
clf.C,
variant='one_slack')
assert_almost_equal(clf.primal_objective_curve_[-1], primal_objective)
# now subgradient. Should also work in batch-mode.
clf = SubgradientSSVM(model=crf, max_iter=5, C=1, batch_size=-1)
clf.fit(X, Y)
primal_objective = objective_primal(clf.model, clf.w, X, Y, clf.C)
assert_almost_equal(clf.objective_curve_[-1], primal_objective)
# frank wolfe
clf = FrankWolfeSSVM(model=crf, max_iter=5, C=1, batch_mode=True)
clf.fit(X, Y)
primal_objective = objective_primal(clf.model, clf.w, X, Y, clf.C)
assert_almost_equal(clf.primal_objective_curve_[-1], primal_objective)
# block-coordinate Frank-Wolfe
clf = FrankWolfeSSVM(model=crf, max_iter=5, C=1, batch_mode=False)
clf.fit(X, Y)
primal_objective = objective_primal(clf.model, clf.w, X, Y, clf.C)
assert_almost_equal(clf.primal_objective_curve_[-1], primal_objective)
def test_objective():
# test that SubgradientLatentSSVM does the same as SubgradientSVM,
# in particular that it has the same loss, if there are no latent states.
X, Y = generate_blocks_multinomial(n_samples=10, noise=.3, seed=1)
inference_method = get_installed(["qpbo", "ad3", "lp"])[0]
n_labels = 3
crfl = LatentGridCRF(n_labels=n_labels, n_states_per_label=1,
inference_method=inference_method)
clfl = SubgradientLatentSSVM(model=crfl, max_iter=20, C=10.,
learning_rate=0.001, momentum=0.98)
crfl.initialize(X, Y)
clfl.w = np.zeros(crfl.size_joint_feature) # this disables random init
clfl.fit(X, Y)
crf = GridCRF(n_states=n_labels, inference_method=inference_method)
clf = SubgradientSSVM(model=crf, max_iter=20, C=10., learning_rate=0.001,
momentum=0.98)
clf.fit(X, Y)
assert_array_almost_equal(clf.w, clfl.w)
assert_almost_equal(clf.objective_curve_[-1], clfl.objective_curve_[-1])
assert_array_equal(clf.predict(X), clfl.predict(X))
assert_array_equal(clf.predict(X), Y)
digits = load_digits()
X, y = digits.data, digits.target
#X = X / 255.
X = X / 16.
#y = y.astype(np.int) - 1
X_train, X_test, y_train, y_test = train_test_split(X, y)
# we add a constant 1 feature for the bias
X_train_bias = np.hstack([X_train, np.ones((X_train.shape[0], 1))])
X_test_bias = np.hstack([X_test, np.ones((X_test.shape[0], 1))])
model = MultiClassClf(n_features=X_train_bias.shape[1], n_classes=10)
n_slack_svm = NSlackSSVM(model, verbose=2, check_constraints=False, C=0.1,
batch_size=100, tol=1e-2)
one_slack_svm = OneSlackSSVM(model, verbose=2, C=.10, tol=.001)
subgradient_svm = SubgradientSSVM(model, C=0.1, learning_rate=0.000001,
max_iter=1000, verbose=0)
fw_bc_svm = FrankWolfeSSVM(model, C=.1, max_iter=50)
fw_batch_svm = FrankWolfeSSVM(model, C=.1, max_iter=50, batch_mode=True)
# n-slack cutting plane ssvm
start = time()
n_slack_svm.fit(X_train_bias, y_train)
time_n_slack_svm = time() - start
y_pred = np.hstack(n_slack_svm.predict(X_test_bias))
print("Score with pystruct n-slack ssvm: %f (took %f seconds)"
% (np.mean(y_pred == y_test), time_n_slack_svm))
## 1-slack cutting plane ssvm
start = time()
one_slack_svm.fit(X_train_bias, y_train)
'branch_and_bound': True
}))
n_slack_svm = NSlackSSVM(crf,
check_constraints=False,
max_iter=50,
batch_size=1,
tol=0.001)
one_slack_svm = OneSlackSSVM(crf,
check_constraints=False,
max_iter=100,
tol=0.001,
inference_cache=50)
subgradient_svm = SubgradientSSVM(crf,
learning_rate=0.001,
max_iter=20,
decay_exponent=0,
momentum=0)
bcfw_svm = FrankWolfeSSVM(crf, max_iter=50, check_dual_every=4)
#n-slack cutting plane ssvm
n_slack_svm.fit(X, Y)
# 1-slack cutting plane ssvm
one_slack_svm.fit(X, Y)
# online subgradient ssvm
subgradient_svm.fit(X, Y)
# Block coordinate Frank-Wolfe
bcfw_svm.fit(X, Y)
X, y = digits.data, digits.target
# make binary task by doing odd vs even numers
y = y % 2
# code as +1 and -1
y = 2 * y - 1
X /= X.max()
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)
pbl = BinaryClf()
n_slack_svm = NSlackSSVM(pbl, C=10, batch_size=-1)
one_slack_svm = OneSlackSSVM(pbl, C=10, tol=0.1)
subgradient_svm = SubgradientSSVM(pbl,
C=10,
learning_rate=0.1,
max_iter=100,
batch_size=10)
# we add a constant 1 feature for the bias
X_train_bias = np.hstack([X_train, np.ones((X_train.shape[0], 1))])
X_test_bias = np.hstack([X_test, np.ones((X_test.shape[0], 1))])
# n-slack cutting plane ssvm
start = time()
n_slack_svm.fit(X_train_bias, y_train)
time_n_slack_svm = time() - start
acc_n_slack = n_slack_svm.score(X_test_bias, y_test)
print("Score with pystruct n-slack ssvm: %f (took %f seconds)" %
(acc_n_slack, time_n_slack_svm))
def make_random_trees(n_samples=50, n_nodes=100, n_states=7, n_features=10):
crf = GraphCRF(inference_method='max-product', n_states=n_states,
n_features=n_features)
weights = np.random.randn(crf.size_joint_feature)
X, y = [], []
for i in range(n_samples):
distances = np.random.randn(n_nodes, n_nodes)
features = np.random.randn(n_nodes, n_features)
tree = minimum_spanning_tree(sparse.csr_matrix(distances))
edges = np.c_[tree.nonzero()]
X.append((features, edges))
y.append(crf.inference(X[-1], weights))
return X, y, weights
X, y, weights = make_random_trees(n_nodes=1000)
X_train, X_test, y_train, y_test = train_test_split(X, y)
#tree_model = MultiLabelClf(edges=tree, inference_method=('ogm', {'alg': 'dyn'}))
tree_model = GraphCRF(inference_method='max-product')
tree_ssvm = SubgradientSSVM(tree_model, max_iter=4, C=1, verbose=10)
print("fitting tree model...")
tree_ssvm.fit(X_train, y_train)
print("Training loss tree model: %f" % tree_ssvm.score(X_train, y_train))
print("Test loss tree model: %f" % tree_ssvm.score(X_test, y_test))
