def test_iris(self):
rca = RCA(dim=2)
chunks = RCA.prepare_constraints(self.iris_labels, num_chunks=30,
chunk_size=2, seed=1234)
rca.fit(self.iris_points, chunks)
csep = class_separation(rca.transform(), self.iris_labels)
self.assertLess(csep, 0.25)
class RCA:
def __init__(self):
self.metric_model = RCA_ml()
self.X_tr = None
self.y_train = None
self.X_te = None
def fit(self, X_tr, y_train):
"""Fits the model to the prescribed data."""
self.X_tr = X_tr
self.y_train = y_train
return self.metric_model.fit(X_tr, y_train)
def transform(self, X):
"""Transforms the test data according to the model"""
return self.metric_model.transform(X)
def predict_proba(self, X_te):
"""Predicts the probabilities of each of the test samples"""
test_samples = X_te.shape[0]
self.X_tr = self.transform(self.X_tr)
clf = NearestCentroid()
clf.fit(self.X_tr, self.y_train)
centroids = clf.centroids_
probabilities = np.zeros((test_samples, centroids.shape[0]))
for sample in xrange(test_samples):
probabilities[sample] = sk_nearest_neighbour_proba(
centroids, X_te[sample, :])
return probabilities
def test_rank_deficient_returns_warning(self):
"""Checks that if the covariance matrix is not invertible, we raise a
warning message advising to use PCA"""
X, y = load_iris(return_X_y=True)
# we make the fourth column a linear combination of the two first,
# so that the covariance matrix will not be invertible:
X[:, 3] = X[:, 0] + 3 * X[:, 1]
rca = RCA()
msg = ('The inner covariance matrix is not invertible, '
'so the transformation matrix may contain Nan values. '
'You should remove any linearly dependent features and/or '
'reduce the dimensionality of your input, '
'for instance using `sklearn.decomposition.PCA` as a '
'preprocessing step.')
with pytest.warns(None) as raised_warnings:
rca.fit(X, y)
assert any(str(w.message) == msg for w in raised_warnings)
def test_iris(self):
rca = RCA(dim=2)
chunks = RCA.prepare_constraints(self.iris_labels,
num_chunks=30,
chunk_size=2,
seed=1234)
rca.fit(self.iris_points, chunks)
csep = class_separation(rca.transform(), self.iris_labels)
self.assertLess(csep, 0.25)
def fit(self, X, y=None, ml=[], cl=[]):
X_transformed = X
if ml:
chunks = np.full(X.shape[0], -1)
ml_graph, cl_graph, neighborhoods = preprocess_constraints(
ml, cl, X.shape[0])
for i, neighborhood in enumerate(neighborhoods):
chunks[neighborhood] = i
# print(chunks)
rca = RCA()
rca.fit(X, chunks=chunks)
X_transformed = rca.transform(X)
# print(rca.metric())
kmeans = KMeans(n_clusters=self.n_clusters, max_iter=self.max_iter)
kmeans.fit(X_transformed)
self.labels_ = kmeans.labels_
return self
ids_quadruplets_learners = list(
map(lambda x: x.__class__.__name__,
[learner for (learner, _) in quadruplets_learners]))
pairs_learners = [
(ITML(), build_pairs),
(MMC(max_iter=2), build_pairs), # max_iter=2 for faster
(SDML(), build_pairs),
]
ids_pairs_learners = list(
map(lambda x: x.__class__.__name__,
[learner for (learner, _) in pairs_learners]))
classifiers = [(Covariance(), build_classification),
(LFDA(), build_classification), (LMNN(), build_classification),
(NCA(), build_classification), (RCA(), build_classification),
(ITML_Supervised(max_iter=5), build_classification),
(LSML_Supervised(), build_classification),
(MMC_Supervised(max_iter=5), build_classification),
(RCA_Supervised(num_chunks=10), build_classification),
(SDML_Supervised(), build_classification)]
ids_classifiers = list(
map(lambda x: x.__class__.__name__,
[learner for (learner, _) in classifiers]))
regressors = [(MLKR(), build_regression)]
ids_regressors = list(
map(lambda x: x.__class__.__name__,
[learner for (learner, _) in regressors]))
WeaklySupervisedClasses = (_PairsClassifierMixin, _QuadrupletsClassifierMixin)
def __init__(self):
self.metric_model = RCA_ml()
self.X_tr = None
self.y_train = None
self.X_te = None
def compute_graph(current_graph=[]):
global image_names, labels, n_clusters
global graph, position_constraints, prev_embedding
global d, features
if len(current_graph) == 0 or prev_embedding is None:
print('Initialise graph...')
tic = time()
compute_embedding() # initialise prev_embedding with standard tsne
# find clusters
clusters = cluster_embedding(prev_embedding,
n_clusters=n_clusters,
seed=seed)
graph = create_graph(image_names,
prev_embedding,
label=clusters,
labels=labels)
toc = time()
print('Done. ({:2.0f}min {:2.1f}s)'.format((toc - tic) / 60,
(toc - tic) % 60))
print('Embedding range: x [{}, {}], y [{}, {}]'.format(
prev_embedding[0].min(), prev_embedding[0].max(),
prev_embedding[1].min(), prev_embedding[1].max()))
return graph
print('Update graph...')
tic = time()
graph = format_graph(current_graph['nodes'])
# get current embedding
current_embedding = prev_embedding.copy()
moved = get_moved(
margin=2.0) # nodes which have moved further than the given margin
if len(moved) > 0:
pos_moved = np.array([[graph[idx]['x'], graph[idx]['y']]
for idx in moved])
current_embedding[moved] = pos_moved
# find clusters
clusters = cluster_embedding(current_embedding,
n_clusters=n_clusters,
seed=seed)
if len(moved) > 0:
# sample chunks from clusters
chunk_size = int(d / 4.99) # minimal chunk size
chunks = make_chunks(current_embedding,
moved,
clusters,
chunk_size,
n_neighbors=5)
# transform features
fts_reduced_rca = RCA().fit_transform(fts_reduced, chunks)
if np.isfinite(fts_reduced_rca).all():
fts_reduced = fts_reduced_rca
else:
warnings.warn(
'RCA features included infinite value or nan, so features are not updated.'
'Try to group more samples or reduce cluster size.',
RuntimeWarning)
compute_embedding() # update prev_embedding
graph = create_graph(image_names,
prev_embedding,
label=clusters,
labels=labels)
print('Embedding range: x [{}, {}], y [{}, {}]'.format(
prev_embedding[0].min(), prev_embedding[0].max(),
prev_embedding[1].min(), prev_embedding[1].max()))
toc = time()
print('Done. ({:2.0f}min {:2.1f}s)'.format((toc - tic) / 60,
(toc - tic) % 60))
return graph
