class LMNNClassifier(BaseEstimator, ClassifierMixin):
def __init__(self, k=3, pca=None, train=True, mu=0.5):
self.k = k
self.train = train
self.pca = pca
self.pca_trasform = None
self.mu = mu
self.lmnn = LMNN(k=k, use_pca=False, max_iter=10000, regularization=mu)
def fit(self, x, y=None):
n, d = x.shape
if self.pca is not None:
pca = PCA(n_components=self.pca)
pca.fit(x)
self.pca_trasform = pca.transform
x = pca.transform(x)
if self.train:
self.lmnn.fit(x, y)
self.knn = KNeighborsClassifier(
n_neighbors=self.k,
metric='mahalanobis',
metric_params=dict(VI=self.lmnn.metric()),
n_jobs=-1)
else:
self.knn = KNeighborsClassifier(n_neighbors=self.k)
self.knn.fit(x, y)
return self
def predict(self, x, y=None):
if self.pca_trasform is not None:
x = self.pca_trasform(x)
return self.knn.predict(x, y)
def score(self, x, y=None):
if self.pca_trasform is not None:
x = self.pca_trasform(x)
return self.knn.score(x, y)
def set_params(self, **parameters):
for parameter, value in parameters.items():
setattr(self, parameter, value)
if parameter == 'mu':
setattr(self.lmnn, 'regularization', value)
return self
def test_lmnn(self): lmnn = LMNN(k=5, learn_rate=1e-6, verbose=False) lmnn.fit(self.X, self.y) L = lmnn.transformer_ assert_array_almost_equal(L.T.dot(L), lmnn.metric())
rank_accuracies_l_2.append(rank_accuracies)
mAP_l_2.append(mAP)
metric_l_2.append('Learnt Mahalanobis (Red. Set)')
# In[24]:
from metric_learn import LMNN
lmnn = LMNN(k=3, learn_rate=1e-6, max_iter=50)
lmnn.fit(X_train_pca, y_train)
M = lmnn.metric()
print ('Metric learnt')
rank_accuracies, mAP = evaluate_metric(X_query_pca, camId_query, y_query,
X_gallery_pca, camId_gallery, y_gallery,
metric ='mahalanobis',
parameters = M)
rank_accuracies_l_2.append(rank_accuracies)
mAP_l_2.append(mAP)
metric_l_2.append('Learnt LMNN')
def test_lmnn(self): lmnn = LMNN(k=5, learn_rate=1e-6, verbose=False) lmnn.fit(self.X, self.y) L = lmnn.transformer() assert_array_almost_equal(L.T.dot(L), lmnn.metric())
